development of motion analysis protocols based on inertial ... - Xsens

PIETRO GAROFALO 

DEVELOPMENT OF 

MOTION ANALYSIS 

PROTOCOLS BASED 

ON INERTIAL SENSORS 

Ph.D. Thesis in Bioengineering

ABSTRACT 

Inertial sensors-<strong>based</strong> systems are relatively recent. Knowledge and 

<strong>development</strong> <strong>of</strong> methods and algorithms for the use <strong>of</strong> such systems for clinical 

purposes is therefore limited, if compared with camera-<strong>based</strong> systems. 

However, their advantages in terms <strong>of</strong> cost effectiveness, portability, small 

size, are valid reasons to follow this direction. 

The <strong>protocols</strong> described in this thesis can be particularly helpful for 

rehabilitation centers in which the high cost <strong>of</strong> instrumentation or limitations in 

the working areas and specialized personnel, do not allow the use <strong>of</strong> camera<strong>based</strong> 

systems. Moreover, many applications requiring upper and lower limb 

<strong>motion</strong> <strong>analysis</strong> to be performed outside the laboratories or when is required 

the active participation <strong>of</strong> the operator while the patient is moving, will benefit 

from these <strong>protocols</strong>. 

The application <strong>of</strong> inertial sensors on lower limb amputees highlights 

conditions which are challenging for magnetomer-<strong>based</strong> systems, due to 

ferromagnetic material commonly adopted for the construction <strong>of</strong> orthopaedic 

devices, idraulic prosthetic components or motors. 

This thesis also describes a solution for solving the above problem by means <strong>of</strong> 

a new method for improving the accuracy <strong>of</strong> the Xsens products in measuring 

3D kinematics. The <strong>development</strong> and validation <strong>of</strong> such technique was carried 

out in the collaboration between Xsens Technologies B.V and the INAIL 

Prostheses Centre (Budrio, Italy). 

In the author’s opinion, this thesis and the <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on 

inertial sensors here described, are a demonstration <strong>of</strong> how a strict 

collaboration between the industry, the clinical centers and the research 

laboratories can improve the knowledge, exchange know-how, with the 

common goal to develop new application-oriented systems. 

1

Alma Mater Studiorum – University <strong>of</strong> Bologna 

DEVELOPMENT OF MOTION 

ANALYSIS PROTOCOLS BASED 


PhD Candidate: 

Eng. Pietro Gar<strong>of</strong>alo 

Tutor: 

Pr<strong>of</strong>. Angelo Cappello 

2 

Co-supervisor: 

Eng. Andrea Giovanni 

Cutti 

Examiner: 

Pr<strong>of</strong>. Ugo Della Croce

DEVELOPMENT OF MOTION 

ANALYSIS PROTOCOLS BASED 


PhD Candidate 

Eng. Pietro Gar<strong>of</strong>alo 

Copyright © 2010 by Pietro Gar<strong>of</strong>alo, Bologna, Italy 

All rights reserved. No part <strong>of</strong> this publication may be reproduced or transmitted in any 

form or by any means, electronic or mechanical, including photocopy, recording or any 

information storage or retrieval system, without permission in writing from the author. 

3

Alla mia famiglia… 

A Silvia… 

A Nino… 

4

CONTENTS 

CHAPTER 1 .................................................................................................................................... 9 

GENERAL INTRODUCTION....................................................................................................... 9 

ABSTRACT .................................................................................................................................... 10 

1.1 CLINICAL AND INSTRUMENTAL MOTION ANALYSIS .................................................. 11 

1.2 MOTION ANALYSIS PROTOCOLS BASED ON INERTIAL SENSORS ............................. 11 

1.3 INAIL PROSTHESES CENTRE .............................................................................................. 20 

1.4 AIM OF THE THESIS AND FRAMEWORK .......................................................................... 23 

1.5 EXPERIENCE AT XSENS TECHNOLOGIES B.V. ................................................................ 26 

1.6 THESIS OUTLINE ................................................................................................................... 28 

1.7 FUNCTIONAL ANATOMY OF THE UPPER-EXTREMITY ................................................. 30 

1.8 SHOULDER PATHOLOGIES AND COMPENSATION STRATEGIES ................................ 43 

1.9 UPPER-EXTREMITY AMPUTATIONS AND PROSTHETIC DEVICES ............................. 51 

1.10 LOWER-EXTREMITY AMPUTATIONS AND PROSTHETIC DEVICES .......................... 57 

1.11 MEASUREMENT SYSTEMS BASED ON INERTIAL AND MAGNETIC SENSORS ....... 70 

1.12 REFERENCES ........................................................................................................................ 82 

CHAPTER 2 .................................................................................................................................. 91 

FUNCTIONAL EVALUATION OF THE UPPER-EXTREMITY THROUGH 

STEREOPHOTOGRAMMETRIC SYSTEMS .......................................................................... 91 

ABSTRACT ................................................................................................................................... 92 

2.1 MOTION ANALYSIS ON NON AMPUTEES ..................................................................... 93 

2.1.1 DEVELOPMENT AND VALIDATION OF A PROTOCOL FOR THE EVALUATION OF THE 

COMPENSATION STRATEGIES IN UPPER-EXTREMITY ...................................................... 93 

2.1.2 APPLICATION SCENARIOS ................................................................................. 114 

2.1.3 REFERENCES .................................................................................................... 129 

2.2 MOTION ANALYSIS ON AMPUTEES ............................................................................. 134 

2.2.1 DEVELOPMENT OF A MOTION ANALYSIS PROTOCOL FOR THE KINEMATICS OF UPPER- 

LIMB MYOELECTRIC PROSTHESES .............................................................................. 134 

2.2.2 REFERENCES .................................................................................................... 149 

2.3 DEVELOPMENT OF THE END-USER CLINICAL SOFTWARE FOR THE UPPER- 

EXTREMITY PROTOCOLS BASED ON STEREOPHOTOGRAMMETRY...................... 151 

2.3.1 UPLIFE - UPPER LIMB FUNCTIONAL EVALUATION TOOLBOX ........................... 151 

5

CHAPTER 3 ................................................................................................................................ 158 

FUNCTIONAL EVALUATION OF THE LOWER-EXTREMITY THROUGH 

STEREOPHOTOGRAMMETRIC SYSTEMS ........................................................................ 158 

ABSTRACT ................................................................................................................................. 158 


3.1.1 DEVELOPMENT OF A PROTOCOL FOR THE EVALUATION OF LOWER-EXTREMITY 

KINEMATICS OF TRANSFEMORAL AMPUTEES .............................................................. 159 


KINETICS OF TRANSFEMORAL AND TRANSTIBIAL AMPUTEES ....................................... 164 

3.1.3 REFERENCES .................................................................................................... 168 

3.2 DEVELOPMENT OF THE END-USER CLINICAL SOFTWARE FOR THE LOWER- 

EXTREMITY PROTOCOLS BASED ON STEREOPHOTOGRAMMETRY...................... 169 

3.2.1 LOLIFE - LOWER LIMB FUNCTIONAL EVALUATION TOOLBOX ......................... 169 

CHAPTER 4 ................................................................................................................................ 177 

FUNCTIONAL EVALUATION OF THE LOWER-EXTREMITY THROUGH INERTIAL 

AND MAGNETIC MEASUREMENT SYSTEMS ................................................................... 177 

ABSTRACT ................................................................................................................................. 177 

4.1 MOTION ANALYSIS ON NON AMPUTEES ................................................................... 178 

4.1.1 OUTWALK PROTOCOL ....................................................................................... 179 

4.1.2 REFERENCES .................................................................................................... 198 


4.2.1 VALIDATION OF OUTWALK PROTOCOL ON BELOW-KNEE AMPUTEES .................. 203 

4.2.2 EVALUATION OF ABOVE-KNEE AMPUTEES KINEMATICS DURING GAIT USING 

INERTIAL SENSORS .................................................................................................... 206 

4.2.3 REFERENCES .................................................................................................... 221 

4.3 DEVELOPMENT OF THE END-USER CLINICAL SOFTWARE FOR THE 

PROTOCOLS BASED ON INERTIAL SENSORS.................................................................. 223 

4.3.1 DESIGN OF OUTWALK MANAGER AND MAIN FEATURES .................................... 224 

4.3.2 USE OF OUTWALK MANAGER IN CLINICAL SETTINGS ........................................ 226 

4.3.3 OUTWALK MANAGER TUTORIAL ...................................................................... 235 

CHAPTER 5 ................................................................................................................................ 240 

FUNCTIONAL EVALUATION OF THE UPPER-EXTREMITY THROUGH INERTIAL 

AND MAGNETIC MEASUREMENT SYSTEMS ................................................................... 240 

ABSTRACT ................................................................................................................................. 240 

5.1 MOTION ANALYSIS ON NON AMPUTEES ................................................................... 241 

5.1.1 DEVELOPMENT OF A PROTOCOL FOR THE EVALUATION OF UPPER-EXTREMITY 

KINEMATICS ............................................................................................................. 241 

6

5.1.2 APPLICATION SCENARIOS ................................................................................. 252 

5.1.3 REFERENCES .................................................................................................... 254 


PROTOCOLS BASED ON INERTIAL SENSORS.................................................................. 255 

5.2.1 IDES MANAGER AND ITS USE IN CLINICAL SETTINGS ........................................ 256 

5.2.2 IDES MANAGER TUTORIAL.............................................................................. 262 

CHAPTER 6 ................................................................................................................................ 267 

A NEW ALGORITHM FOR THE APPLICATION ON AMPUTEES OF THE LOWER 

AND UPPER-EXTREMITY PROTOCOLS BASED ON INERTIAL SENSORS ................ 267 

6.1 INTRODUCTION ................................................................................................................... 268 

6.2 KIC (KINEMATIC COUPLING) ALGORITHM ................................................................... 271 

6.3 INTERFACING KIC ALGORITHM WITH UPPER AND LOWER-EXTREMITY 

PROTOCOLS ............................................................................................................................... 274 

6.4 REFERENCES ........................................................................................................................ 281 

CHAPTER 7 ................................................................................................................................ 283 

DATA VARIABILITY IN MOTION ANALYSIS .................................................................... 283 

ABSTRACT .................................................................................................................................. 283 

7.1 SEGMENTATION OF MOVEMENT .................................................................................... 284 

7.2 REFERENCES ........................................................................................................................ 290 

CHAPTER 8 ................................................................................................................................ 292 

CONCLUSIONS.......................................................................................................................... 292 

CHAPTER 9 ................................................................................................................................ 299 

PUBLICATIONS ........................................................................................................................ 299 

ABOUT THE AUTHOR ............................................................................................................. 305 

RINGRAZIAMENTI .................................................................................................................. 310 

ACKNOWLEDGMENTS ........................................................................................................... 314 

7

CHAPTER 1 

GENERAL INTRODUCTION 

ABSTRACT 

1.1 CLINICAL AND INSTRUMENTAL MOTION ANALYSIS 

1.2 MOTION ANALYSIS PROTOCOLS BASED ON INERTIAL SENSORS 

1.3 INAIL PROSTHESES CENTRE 

1.4 AIM OF THE THESIS AND FRAMEWORK 

1.5 EXPERIENCE AT XSENS TECHNOLOGIES B.V. 

1.6 THESIS OUTLINE 

1.7 FUNCTIONAL ANATOMY OF THE UPPER-EXTREMITY 

1.8 SHOULDER PATHOLOGIES AND COMPENSATION STRATEGIES 

1.9 UPPER-EXTREMITY AMPUTATIONS AND PROSTHETIC DEVICES 

1.10 LOWER-EXTREMITY AMPUTATIONS AND PROSTHETIC DEVICES 

1.11 MEASUREMENT SYSTEMS BASED ON INERTIAL AND MAGNETIC SENSORS 

1.12 REFERENCES 

9

ABSTRACT 

The aim <strong>of</strong> this thesis was to describe the <strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> 

<strong>protocols</strong> for applications on upper and lower limb extremities, by using inertial 

sensors-<strong>based</strong> systems (IMMS). IMMS are relatively recent. Knowledge and 

<strong>development</strong> <strong>of</strong> methods and algorithms for the use <strong>of</strong> such systems for clinical 

purposes is therefore limited if compared with stereophotogrammetry. 

However, their advantages in terms <strong>of</strong> low cost, portability, small size, are a 

valid reason to follow this direction. 

When developing <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on IMMs, attention must be 

given to several aspects, like the accuracy <strong>of</strong> inertial sensors-<strong>based</strong> systems and 

their reliability. The need to develop specific algorithms/methods and s<strong>of</strong>tware 

for using these systems for specific applications, is as much important as the 

<strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on them. 

For this reason, the goal was achieved first <strong>of</strong> all trying to correctly design the 

<strong>protocols</strong> <strong>based</strong> on IMMS, in terms <strong>of</strong> exploring and developing which features 

were suitable for their specific application. The use <strong>of</strong> optoelectronic systems 

was necessary because they provided a gold standard and accurate 

measurement, which was used as a reference for the validation <strong>of</strong> the <strong>protocols</strong> 

<strong>based</strong> on IMMS. 

Therefore this thesis will describe the <strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> 

<strong>protocols</strong> <strong>based</strong> on IMMS, for clinical applications on upper and lower 

extremities pathologies, starting from the gold standard <strong>motion</strong> <strong>analysis</strong> 

performed through the optoelectronic systems, adopting and developing 

common methodologies in terms <strong>of</strong> methods for the <strong>protocols</strong> validation, data 

<strong>analysis</strong>, algorithms and end-user clinical s<strong>of</strong>tware. 

10

1.1 Clinical and instrumental <strong>motion</strong> <strong>analysis</strong> 

In the contemporary medicine the patient is the starting and ending point <strong>of</strong> a 

circular path [1]. Practitioners are directly in contact with the patient but at the 

same time instrumentation as support to the diagnosis and/or the therapy is 

adopted. Instrumental <strong>analysis</strong> can be adopted by the practitioners in order to 

allow them to mostly concentrate on the therapy decision-making process and 

to improve the knowledge about a specific biological system. In fact, without 

the use <strong>of</strong> instrumental <strong>analysis</strong>, for example in the case <strong>of</strong> the musculo-skeletal 

system, physicians are not able to deeply examine the biological systems from 

the anatomical and physiological points <strong>of</strong> view. 

Bioengineers play the role <strong>of</strong> designing the <strong>analysis</strong> tools required from the 

practitioner for the examination <strong>of</strong> the biological system. An active 

collaboration between bioengineers and practitioners is necessary in order to 

provide the engineer with the right information about the clinical question to 

solve and, from the other side, to provide the practitioner with the necessary 

knowledge about the optimal way <strong>of</strong> using the technology. The latter does not 

include only the way in which a device should be correctly used, but also the 

way in which all the benefits coming from its use can be understood. 

The application <strong>of</strong> instrumental <strong>analysis</strong> on the rehabilitation field is <strong>based</strong> on 

several aspects, like the instrumentation adopted, the mathematical models, the 

algorithms, the data processing. The combination <strong>of</strong> these elements determines 

the complexity <strong>of</strong> the <strong>analysis</strong> system and at the same time its validity. 

As it will be discussed later, it is worth to say that a general purpose system is 

not necessarily the best choice for supporting the clinical routine examinations. 

In fact the characteristics <strong>of</strong> an <strong>analysis</strong> system have to be close to the one in 

clinical settings. For example the time required for performing a clinical 

examination using a <strong>motion</strong> <strong>analysis</strong> system has to be similar to the one spent 

during a normal routine examination or, even, the time required for the <strong>motion</strong> 

<strong>analysis</strong> system must be less than that, when the system is adopted as additional 

instrument together with clinical evaluation scales. 

1.2 Motion <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on inertial sensors 

Instrumental <strong>motion</strong> <strong>analysis</strong> has been widely adopted in clinics for upper and 

lower extremity functional assessment. 

Lower limb gait <strong>analysis</strong> has been the main application area <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> 

11

<strong>protocols</strong>, developed since the ‗60s. 

More recently, <strong>protocols</strong> and specific instrumentation were developed for the 

upper limb functional evaluation. Their spread among the research and clinical 

laboratories was supported by the limitations in the so called ―visual 

observation‖. Without objective measurements and with low sensitivity, this 

discipline could lead to misinterpretation <strong>of</strong> the results or errors in the 

diagnostic process [2]. 

1.2.1 The technology 

As in other fields <strong>of</strong> engineering, the more the improvements in the technology, 

the more the <strong>motion</strong> capture techniques have become refined. 

Motion tracking systems <strong>based</strong> on stereophotogrammetry provide high 

accuracy in tracking the position <strong>of</strong> markers [3-6]. Although measurements 

through stereophotogrammetry can be optimized in terms <strong>of</strong> time and 

resources, this kind <strong>of</strong> <strong>analysis</strong> is strictly depending on the use <strong>of</strong> cameras 

which restricts their use in the laboratory workspace to work properly; the 

instrumentation is typically expensive; the installation is difficult in small 

<strong>of</strong>fices where the <strong>analysis</strong> is needed. Data <strong>analysis</strong> requires specialized 

personnel, it is time consuming and with some limitations in terms <strong>of</strong> 

application. For instance, due to the necessity <strong>of</strong> marker visibility (in the case 

<strong>of</strong> passive markers are used) or problems <strong>of</strong> cabling (in the case <strong>of</strong> active 

markers are used), measurements during mobilization (Chapter 5) are far to be 

conducted using cameras, which means that the practitioners are not allowed to 

act on the patient during <strong>motion</strong> <strong>analysis</strong>. Recently, real-time results <strong>of</strong> the 

<strong>analysis</strong>, providing feedback to the user, are available (e.g. Vicon Nexus 

s<strong>of</strong>tware [7]), but the procedures are typically complex and restricted to a small 

ensemble <strong>of</strong> clinical parameters. 

Despite <strong>of</strong> the above limitations, systems <strong>based</strong> on stereophotogrammetry have 

been considered the golden standard in <strong>motion</strong> <strong>analysis</strong>. The technology at the 

base <strong>of</strong> these systems characterize them as products in the market <strong>of</strong> <strong>motion</strong> 

capture systems, called ―continuous innovations‖ [8], referring to the fact that 

the evolution and upgrades in the field <strong>of</strong> <strong>motion</strong> capture through 

stereophotogrammetry do not require us to change the way in which we use 

them, i.e. the starting point for <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> using 

stereophotogrammetry are marker trajectories and most <strong>of</strong> the <strong>protocols</strong> are 

normally <strong>based</strong> on them, despite <strong>of</strong> the innovations in terms <strong>of</strong> real-time 

algorithms for the trajectory estimation or its filtering. The technology adopted 

12

in the stereophotogrammetric systems provides information that can be 

translated into clinical meaning adopting the methodology developed in the 

past 50 years. 

The advent <strong>of</strong> MEMS (Micro-Electro-Mechanical Systems) technology allowed 

systems <strong>based</strong> on inertial and magnetic sensors (also called Inertial and 

Magnetic Measurement System, IMMS) to be introduced in the biomedical 

community first as additional tool per specific applications (Parkinson‘s disease 

evaluation using accelerometers [9]), then as systems potentially useful for 

ambulatory measurements (activity monitoring [10]), overcoming some <strong>of</strong> the 

limitations described above. Xbus kit <strong>based</strong> on MTx (Xsens Technologies B.V., 

The Netherlands) [11], represented in Figure 1 is an example <strong>of</strong> wearable 

<strong>motion</strong> <strong>analysis</strong> system which includes inertial and magnetic sensors. The 

system has small dimensions; it is completely portable and low cost with 

respect to camera-<strong>based</strong> systems. 

Figure 1 –Xbus kit from Xsens Technologies B.V. 

13

Figure 2 shows the working principle at the base <strong>of</strong> the MTx unit. Each MTx 

unit contains a 3D accelerometer, 3D rate <strong>of</strong> turn (gyroscope) and a 3D 

magnetometer. By fusing the information <strong>of</strong> the three different sensors with a 

Kalman filter-<strong>based</strong> algorithm (XKF3) the 3D orientation <strong>of</strong> the MTx casing 

with respect to a global coordinate system is provided. The global coordinate 

system is created using information from the inertial and magnetic sensors. 

Figure 2 – working principle <strong>of</strong> MTx inertial and magnetic measurement unit 

The system described above is not able to provide an estimation <strong>of</strong> the position 

<strong>of</strong> the MTx in space, due to vulnerability to integration drifts, which does not 

permit an accurate estimation <strong>of</strong> external or internal anatomical landmarks 

position. Moreover, the presence <strong>of</strong> magnetometers can be a limitation when 

the magnetic field in the environment becomes non homogeneous, although the 

current fusion algorithms are robust to rapid and strong variations <strong>of</strong> the 

magnetic field. 

As this thesis will describe, magnetic distortions due to the presence <strong>of</strong> 

ferromagnetic materials inside <strong>of</strong> limb prostheses do not allow the use <strong>of</strong> 

IMMS in certain conditions, unless special techniques which will be presented. 

Recently, only adopting specific techniques and other systems in parallel it is 

possible to have a good estimate <strong>of</strong> position [12]. 

As the useful information provided by each MTx unit, adopting the Xbus kit as 

a ubiquitous system, is its orientation in space, the starting point for the 

<strong>development</strong> <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on MTx are not marker 

trajectories. Furthermore there can be some limitations about the environment 

in which the <strong>analysis</strong> can be performed. 

The attitude <strong>of</strong> the biomedical community toward MEMS technology, changes 

completely with respect to stereophotogrammetry. IMMS open a window into 

the new concept <strong>of</strong> ambulatory <strong>motion</strong> <strong>analysis</strong> but at the same time the 

evolution and upgrades in the field <strong>of</strong> <strong>motion</strong> capture through inertial sensors 

characterize theme as ―discontinuous innovations‖ [8]. In fact, in this case, the 

14

starting point <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> using inertial and magnetic sensors 

are not marker trajectories anymore, rather accelerations, angular velocities, 

magnetic field, orientation in space. 

Therefore, from one side the high accuracy in tracking the position <strong>of</strong> markers 

does not directly imply an effective <strong>motion</strong> <strong>analysis</strong>, being the protocol adopted 

and the manual data processing behind playing an important part in it, and due 

to the limitations <strong>of</strong> stereophotogrammetry, some clinical applications are not 

possible. 

On the other side, the <strong>development</strong> <strong>of</strong> portable systems <strong>based</strong> on new MEMS 

technology, potentially allowing ambulatory <strong>motion</strong> <strong>analysis</strong> and simplifying 

the data processing, does not imply high accuracy in <strong>motion</strong> tracking and 

moreover, to be suitable in clinical settings. Again, in fact the protocol to adopt 

plays an important role and the specifications for the design <strong>of</strong> <strong>protocols</strong> for 

systems <strong>based</strong> on inertial and magnetic sensors change radically. In other 

words, while for the biomedical community it is common to adopt 

stereophotogrammetry in gait <strong>analysis</strong>, the way in which this can be done using 

inertial and magnetic sensors partially needs to be discovered and the relative 

knowledge spread among the community. 

1.2.2 Design <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol 

A <strong>motion</strong> <strong>analysis</strong> protocol is required next to the <strong>motion</strong> capture system, being 

either <strong>based</strong> on stereophotogrammetry or inertial sensors. Sometimes, <strong>motion</strong> 

<strong>analysis</strong> <strong>protocols</strong> are created around the instrumentation available at the 

moment. This implies that some <strong>of</strong> the features contained in these <strong>protocols</strong> are 

not specifications <strong>of</strong> a design, therefore not strictly related to their application. 

As stated in Kontaxis et al [13], a <strong>motion</strong> <strong>analysis</strong> protocol is ―[..] the means <strong>of</strong> 

measurement <strong>of</strong> the parameters required to test the hypotheses at the base <strong>of</strong> 

the study research question‖. This definition implies that the <strong>development</strong> <strong>of</strong> 

<strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> should start from the aim <strong>of</strong> its application, that is the 

research or clinical question. The better the protocol is designed, the more its 

effectiveness. Note that the above definition does consider the instrumentation 

adopted as part <strong>of</strong> it, not out <strong>of</strong> it. Therefore, the choice <strong>of</strong> the instrumentation 

should be also taken into account when developing <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong>, 

especially when the instrumentations available are <strong>based</strong> on different 

technologies. 

Although the specifications <strong>of</strong> objectivity and repeatability might be satisfied, 

15

instrumental <strong>motion</strong> <strong>analysis</strong> does not always meet the requirements <strong>of</strong> a low 

cost, and not time-consuming muscolo-skeletal evaluation. This can lead to not 

negligible disadvantages from the clinical point <strong>of</strong> view. 

In order to improve the efficacy and efficiency <strong>of</strong> the clinical evaluation <strong>based</strong> 

on instrumental <strong>analysis</strong>, it is necessary to find an instrumental tool which can 

be suitable in clinical and rehabilitation environments, especially when few 

resources are available. 

The instrumental tool must be defined in relation with the particular impairment 

as object <strong>of</strong> the evaluation and the main aim to be achieved. The specifications 

by which a <strong>motion</strong> <strong>analysis</strong> protocol must be selected or created correspond to 

some constraints in the clinical settings. Therefore, when designing a protocol 

for a specific clinical quest, the following elements must be considered: 

 

 

 

 

 

the objectives <strong>of</strong> the <strong>analysis</strong> and the specifications <strong>of</strong> clinical 

procedures; 

the constraints and the limitations due to the environment and the 

biological system object <strong>of</strong> the study; 

the instrumentation available in the market and/or the laboratory; 

the numerical quantities to be extracted from the system and necessary 

for its description; 

the error quantities which have to be considered during the final 

interpretation <strong>of</strong> the data 

Using a mathematical language we can assert that the choice <strong>of</strong> a <strong>motion</strong> 

<strong>analysis</strong> protocol corresponds to the determination <strong>of</strong> a function (typically not 

linear) depending on a certain number <strong>of</strong> parameters which coincides with the 

degrees <strong>of</strong> freedom <strong>of</strong> the protocol (the choice <strong>of</strong> the number <strong>of</strong> markers to be 

adopted or the duration <strong>of</strong> the <strong>analysis</strong>, for instance). The variability <strong>of</strong> these 

parameters corresponds to the boundary conditions <strong>of</strong> the mathematical 

problem and the range <strong>of</strong> values the parameters can have is defined by the 

bioengineer in collaboration with the practitioner, case by case. 

The solution <strong>of</strong> this mathematical problem is therefore the function such as all 

the parameters determining it will have values among the range <strong>of</strong> variability 

permitted. 

When no solution is found, it means that some <strong>of</strong> the constraints <strong>of</strong> the design 

are not satisfied and two different scenarios can be considered: 

1) In the first scenario the protocol selected provides an overestimation <strong>of</strong> the 

16

quantitative information required as output <strong>of</strong> its application. This is the case <strong>of</strong> 

a protocol not easily applicable in clinical settings, for example a protocol with 

a great amount <strong>of</strong> data as output providing results too difficult to be interpreted 

and adopted for diagnostic or therapeutic purposes. In this case, all the 

resources adopted for obtaining results with high accuracy do not increase the 

benefit-cost ratio. 

2) The second scenario is opposite to the first. The values assumed by the 

parameters lead to an underestimation <strong>of</strong> the information required as output <strong>of</strong> 

the application. As a consequence the results can be incorrect and with low 

applicability in the clinical diagnosis or clinical decision-making process. 

When a solution is found, the protocol is suitable for the specific instrumental 

<strong>analysis</strong>, although the solution is not necessarily unique. 

In summary, the design <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol must take into account 

the available instrumentation/technology, the specifications deriving from the 

research/clinical question and elements supporting its reliability and 

effectiveness. 

1.2.3 Design <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on inertial sensors 

In the case <strong>of</strong> MEMS technology, in particular IMMS, their combined use by 

means <strong>of</strong> optimization algorithms (Chapter 6), allow the measurement <strong>of</strong> the 

orientation <strong>of</strong> body segments and, through specific techniques [12], the 

measurement <strong>of</strong> their position, although without the combined use <strong>of</strong> other 

instrumentations the accuracy is not comparable with the one <strong>of</strong> 

stereophotogrammetric systems. The manual intervention <strong>of</strong> the user is 

therefore shifted from the data processing <strong>of</strong> marker trajectories, with 

stereophotogrammetry, to the next steps: to validate the accuracy in the 

estimation <strong>of</strong> position and orientation and to extract valuable and meaningful 

clinical information from repeatable and reliable measurements. 

These steps are required when using all kind <strong>of</strong> technology. However, being 

<strong>motion</strong> capture systems <strong>based</strong> on inertial sensors more recent than the ones 

<strong>based</strong> on stereophotogrammetry, the knowledge available about methods to 

perform the above steps is limited when comparing it with golden standard 

systems. 

The positioning <strong>of</strong> markers on the body segments and s<strong>of</strong>t tissue artefact 

reduction techniques are example <strong>of</strong> methodologies developed during the past 

17

years using stereophotogrammetric systems and that can be applied when 

developing a <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on IMMS. 

There are several aspects to consider when developing <strong>motion</strong> capture systems 

<strong>based</strong> on IMMS. 

Firstly, not all the methodology developed in the past years using systems 

<strong>based</strong> on stereophotogrammetry, can be directly applied when using IMMS. 

For instance, anatomical calibrations <strong>based</strong> on anatomical landmark palpation 

are not possible when the system cannot provide an accurate estimation <strong>of</strong> 

external landmarks. 

Secondly, the <strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on IMMS 

naturally walks in parallel with the main advantage <strong>of</strong> having a portable 

system: to provide a ubiquitous <strong>motion</strong> <strong>analysis</strong> system. This augments the 

number <strong>of</strong> protocol specifications when using portable systems, with 

requirements different than stereophotogrammetric systems ones. This means 

that some methodologies which are required when using systems <strong>based</strong> on 

inertial and magnetic sensors can be developed and validated using the 

stereophotogrammetric systems (e.g. functional methods), but the evaluation <strong>of</strong> 

a portable system, providing an ubiquitous <strong>analysis</strong>, is not always possible 

inside <strong>of</strong> a laboratory. 

In general we can assert that <strong>protocols</strong> <strong>based</strong> on IMMS are the tool by which 

new methodologies developed for being compatible with the new technology, 

are applied to specific questions. 

Furthermore, the correct design <strong>of</strong> the protocol is not the unique step to perform 

when creating a <strong>motion</strong> <strong>analysis</strong> system. In fact, in order to provide quantitative 

measures for supporting the clinicians in better understanding pathology or in 

the decision-making process, all the measurements must be repeatable and 

validated. The last two are indeed important requirements for all the 

methodologies and <strong>protocols</strong> developed, in order to be shared among the 

community and finally adopted in practice. 

Starting from the above considerations it is useful to consider the golden 

standard systems as the starting point for the <strong>development</strong> and improvement <strong>of</strong> 

<strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on IMMS. 

From one side, there is the need to apply valid and reliable methods using 

18

stereophotogrammetry for evaluating the accuracy and precision <strong>of</strong> systems 

<strong>based</strong> on new technology (such as MEMS inertial and magnetic sensors). 

From the other side, novel methodologies must be developed for using IMMS, 

due to specifications in the protocol design which do not come from the 

experience with stereophotogrammetry. 

Finally, the validation <strong>of</strong> <strong>protocols</strong> <strong>based</strong> on IMMS, in terms <strong>of</strong> their capability 

in providing clinically meaningful information, is a needful step for the 

introduction <strong>of</strong> these tools into the biomedical and clinical community. 

19

1.3 INAIL Prostheses Centre 

The clinical questions at INAIL Prostheses Centre were the main motivation at 

the base <strong>of</strong> this thesis. INAIL is the Italian workers‘ compensation authority, 

―not just compensation but a global protection system for all workers‖. 

INAIL has three main goals: 

1) the prevention <strong>of</strong> work-related injuries; 

2) to provide the injured workers with adequate medical treatments and 

economic benefits, 

3) to return them to active and participated social and working life. 

Table 1 - INAIL statistics about work-related injuries in Italy 

Table 1 presents a recent summary (DATI INAIL, number 11, 2009) <strong>of</strong> workrelated 

injuries (including injuries occurring while moving from house to work 

and vice versa) occurred during the last half <strong>of</strong> 2008 and the first half <strong>of</strong> 2009. 

Most <strong>of</strong> the injuries occur during ordinary work (which means not during work 

which takes place outside the building). Although decreasing from 2008 to 

2009, the number <strong>of</strong> injuries is still impressive. 

INAIL services are provided through the work <strong>of</strong> 235 local <strong>of</strong>fices on the 

Italian territory. 

In particular, INAIL Prostheses Centre provides INAIL services as centre for 

the experimentation and application <strong>of</strong> prostheses and orthopaedic devices, by 

20

searching for new technologies to be introduced in the production <strong>of</strong> 

components, by producing and providing patients with new prostheses and 

orthopaedic devices, by controlling the steps <strong>of</strong> the rehabilitation process. 

The criteria for the work reintegration are <strong>based</strong> on the correct assessment <strong>of</strong> 

the level <strong>of</strong> disability and the working potentiality. From the economical and 

social points <strong>of</strong> view, in order to achieve the above goals, the main objectives 

are to quantitatively and objectively evaluate the impairment <strong>of</strong> patients with 

work-related injuries, the ability in controlling their prosthetic devices and the 

<strong>development</strong> <strong>of</strong> new prosthetic 

devices. 

At INAIL Prostheses Centre the 

rehabilitation process is <strong>based</strong> on 

a multi-disciplinary team 

approach to solve the clinical 

question [ 14 ]. 

These objectives can be reached 

through systems that support the 

physiotherapist in the clinical 

decisions and guide him during 

the rehabilitation. 

In general, clinicians need to be 

supported in the same place the 

rehabilitation treatment is 

Figure 3 – INAIL Prostheses Centre 

Vigorso di Budrio (BO), Italy 

21 

performed, with fast and accurate 

measurements with ambulatory 

systems, fast and reliable 

<strong>protocols</strong> for measuring the mobility and the activities <strong>of</strong> the daily living. 

When these kinds <strong>of</strong> systems will be available, INAIL local <strong>of</strong>fices will be able 

to monitor the evolution <strong>of</strong> patients‘ motor ability over time and the 

orthopaedic technicians will evaluate whether the patient is correctly learning 

how to control his own prosthesis. 

As previously described, despite <strong>of</strong> the good results in terms <strong>of</strong> measurement 

accuracy and applicability, <strong>protocols</strong> <strong>based</strong> on stereophotogrammetry cannot be 

easily adopted in the clinical settings above described. A possible solution to 

these limitations comes from IMMS. 

In order to achieve the above objectives the <strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> 

<strong>protocols</strong> <strong>based</strong> on IMMs, overcoming the limitations <strong>of</strong> the systems <strong>based</strong> on

stereophotogrammetry, required that: 

1) the <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on stereophotogrammetry, already 

adopted at INAIL Prostheses Centre, were improved for being applied 

to specific applications, such as shoulder pathologies; 

2) to start multicenter trials on the application <strong>of</strong> the above <strong>protocols</strong> for 

exploring their validity not only in the research field but also in the 

routine clinical examination. 

3) to develop new methodologies, algorithms and s<strong>of</strong>tware for supporting 

the validation <strong>of</strong> the <strong>protocols</strong> <strong>based</strong> on stereophotogrammetry, taking 

in consideration the later need <strong>of</strong> validating <strong>protocols</strong> <strong>based</strong> on inertial 

sensors. 

INAIL Prostheses Centre was provided with a Vicon MX4 system with 6 

infrared cameras (Oxford Metrics, UK), two force plates (Kistler Instrumente 

AG, Switzerland), as instrumentation for upper and lower limb <strong>motion</strong> <strong>analysis</strong> 

using stereophotogrammetry. 

The inertial and magnetic measurement system adopted for ambulatory <strong>analysis</strong> 

was the Xbus Kit <strong>based</strong> on MTx, previously cited. Further description <strong>of</strong> this 

system can be found in the next sections. 

22

1.4 Aim <strong>of</strong> the thesis and framework 

Starting from the INAIL needs previously described, the main goal <strong>of</strong> this 

thesis was the <strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on IMMS, for 

upper and lower extremities application. 

This goal was achieved from one side trying to correctly design the <strong>protocols</strong> 

<strong>based</strong> on inertial sensors, in terms <strong>of</strong> exploring and developing which features 

were suitable for the specific application <strong>of</strong> the <strong>protocols</strong>. 

From the other side, despite <strong>of</strong> the limitations <strong>of</strong> the optoelectronic systems, 

their use was necessary because: 

1) they provided a gold standard and accurate measurement, which was used as 

reference for the validation <strong>of</strong> the <strong>protocols</strong> <strong>based</strong> on inertial sensors. 

2) all the basic knowledge needful for the creation <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol 

(e.g. s<strong>of</strong>t tissue artefact, positioning <strong>of</strong> markers), created by the community in 

the past 50 years, was the starting point for the generation <strong>of</strong> the same kind <strong>of</strong> 

knowledge in the case <strong>of</strong> inertial sensors. 

As this thesis will demonstrate, the two worlds <strong>of</strong> instrumentation, systems 

<strong>based</strong> on stereophotogrammetry on one side, and inertial and magnetic sensors 

on the other side, were able to exist side by side and moreover this thesis will 

show how <strong>protocols</strong> developed starting from these instrumentations can be 

complementary, depending on the clinical application. 

Therefore this thesis will describe the <strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> 

<strong>protocols</strong>, <strong>based</strong> on inertial sensors, for clinical applications on upper and lower 

extremities pathologies (hereinafter called UX and LX), starting from the gold 

standard <strong>motion</strong> <strong>analysis</strong> performed through the optoelectronic systems, 

adopting and developing common methodologies in terms <strong>of</strong> methods for the 

<strong>protocols</strong> validation, data <strong>analysis</strong>, algorithms and s<strong>of</strong>tware. Moreover, the 

<strong>development</strong> <strong>of</strong> <strong>protocols</strong> must take into account the fact that, for each 

application, <strong>protocols</strong> <strong>based</strong> on inertial and magnetic sensors and <strong>protocols</strong> 

<strong>based</strong> on stereophotogrammetry should be able to measure the same level <strong>of</strong> 

impairment. 

This thesis was almost entirely carried out at the INAIL Prostheses Centre, with 

the collaboration <strong>of</strong> the Department <strong>of</strong> Electronics, Computer Science and 

Systems (Bologna, Italy) and Xsens Technologies B.V. (Enschede, The 

23

Netherlands). 

In order to achieve the goal described above, the main objectives were: 

1) to support the creation and to perform the validation <strong>of</strong> a protocol 

<strong>based</strong> on inertial sensors for patients with shoulder pathologies like 

rotator-cuff tears and shoulder instability; 

2) to support the creation and validation <strong>of</strong> a gait <strong>analysis</strong> protocol <strong>based</strong> 

on inertial sensors specifically designed for amputees; 

3) to validate the use <strong>of</strong> a new method for applying the protocol in 1) to 

subjects with lower limb amputation; 

4) to develop s<strong>of</strong>tware for supporting the validation <strong>of</strong> the <strong>protocols</strong> 

<strong>based</strong> on inertial sensors and to transform them into end-user clinical 

s<strong>of</strong>tware which can simplify the application <strong>of</strong> all the above <strong>protocols</strong> 

in clinical settings. 

The framework adopted during the 3-years research is schematized in Figure 4. 

Figure 4 – Framework adopted during the <strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on 

inertial sensors 

24

During the first year <strong>of</strong> research, the <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on 

stereophotogrammetry, developed for the functional assessment <strong>of</strong> UX on 

patients with shoulder pathologies, was completed and extended for the 

applications on trans-humeral amputees. This work was published in [15]. 

Moreover, the need to study LX kinematics <strong>of</strong> amputees during gait required to 

provide INAIL with the s<strong>of</strong>tware tools for the application <strong>of</strong> the CAST [16] 

technique and the <strong>development</strong> <strong>of</strong> a specific protocol for transfemoral 

amputees. This part was already presented during conferences in 2009 and it 

was included in the full paper submitted to an international journal [17] . 

UX and LX studies shared CAST, algorithms and s<strong>of</strong>tware (UpLiFE and 

LoLiFE toolboxes). In the same year, some preliminary studies on UX and LX 

<strong>protocols</strong> <strong>based</strong> on inertial sensors were performed. 

During the second year, the intra and inter operator repeatability study <strong>of</strong> the 

protocol <strong>based</strong> on stereophotogrammetry for applications on UX was carried 

out [18] and the necessary data required for the repeatability study <strong>of</strong> the 

protocol <strong>based</strong> on IMMS for application on scapula tracking and on lower limb 

amputees were acquired. 

The methodology developed for the <strong>protocols</strong> <strong>based</strong> on stereophotogrammetry 

was adopted for the experiments using inertial sensors. During the last year <strong>of</strong> 

research, the validation <strong>of</strong> the <strong>protocols</strong> <strong>based</strong> on IMMS was carried out for 

applications on UX [19], and during the experience abroad at Xsens 

Technologies B.V. a new methodology for developing the protocol for aboveknee 

amputees kinematics through the MTx system was tested and validated. 

This work was submitted to the ISPO 2010 conference (Leipzig, Germany) and 

was accepted as oral presentation. A full paper is going to be finalized. 

The final step for the <strong>protocols</strong> developed during the past years was to create 

the end-user clinical s<strong>of</strong>tware for their application in clinical settings [20] . 

Finally, the protocol <strong>based</strong> on stereophotogrammetry for the evaluation <strong>of</strong> 

amputees‘ kinematics was adopted together with a kinetic model, in order to 

compare the calculation <strong>of</strong> lower limb 3D joint moments in gait <strong>analysis</strong> 

obtained through two alternative methods: inverse dynamics (<strong>based</strong> on Newton- 

Euler mechanics) and Ground Reaction Force Vector approach [21]. Two case 

studies on transfemoral and transtibial amputees were reported [22]. 

25

1.5 Experience at Xsens Technologies B.V. 

Figure 5 – Xsens 

Technologies B.V. 

The last year <strong>of</strong> research was almost entirely spent at 

Xsens Technologies B.V. (Enschede The 

Netherlands). Founded in 2000, Xsens Technologies is 

market leader in miniature 3D tracking <strong>based</strong> on 

MEMS technology. By their experience and unique 

intellectual property in the field <strong>of</strong> multi-sensor data 

fusion algorithms, Xsens produces innovative products 

which have been also adopted during the past years for 

applied research in the movement science field [23- 

26]. Xsens Technologies is the producer <strong>of</strong> the IMMS 

<strong>based</strong> on MTx adopted in the experiments for the 

<strong>protocols</strong> <strong>based</strong> on inertial sensors at INAIL Prostheses 

Centre. 

The long experience at Xsens Technology was necessary in order to: 

1) improve the knowledge about IMMs and related algorithms for estimating 

orientation in space; 

2) extend the UX and LX <strong>protocols</strong> <strong>based</strong> on IMMS, validated for applications 

on healthy subjects so that they could be applied on amputees, where the 

presence <strong>of</strong> ferromagnetic materials influenced the measurement and studying 

in which conditions they can result critical; 

3) improve the s<strong>of</strong>tware developed at INAIL Prostheses Centre, according to 

the new algorithms and capabilities <strong>of</strong>fered by the S<strong>of</strong>tware Development Kit 

by Xsens; 

4) simplify the execution <strong>of</strong> the experiments aiming to validate the <strong>protocols</strong>, in 

terms <strong>of</strong> synchronization between different measurement systems and spotchecks 

on the quality <strong>of</strong> the data provided by the IMMS; 

5) understand how the <strong>protocols</strong> developed at INAIL Prostheses Centre could 

be optimized for specific UX and LX applications, such as gait <strong>analysis</strong> <strong>of</strong> 

transfemoral and transtibial amputees, in clinical settings; 

6) work in contact with the industrial environment in connection with dutch and 

26

italian researchers for creating clinical applications <strong>of</strong> IMMS. 

27

1.6 Thesis outline 

In the research project both optoelectronic systems and inertial sensors were 

adopted as instrumentation and both UX and LX were considered as targets <strong>of</strong> 

the <strong>protocols</strong>. Common methodology is applied depending on the specific 

application. 

For this reason the thesis is divided into 5 main areas. The general introduction 

(Chapter 1), UX (Chapter 2) and LX (Chapter 3) <strong>protocols</strong> <strong>based</strong> on 

stereophotogrammetry, LX (Chapter 4) and UX (Chapter 5) <strong>protocols</strong> <strong>based</strong> on 

IMMS. Each area is further divided into two parts depending on the kind <strong>of</strong> 

subject which was the target <strong>of</strong> the protocol design (Non amputee subjects and 

amputees). Each part contains specific paragraphs related to the <strong>development</strong> <strong>of</strong> 

the <strong>protocols</strong> and/or their validation. Some application scenarios per each area 

are also reported. Some paragraphs are dedicated to the description <strong>of</strong> the 

s<strong>of</strong>tware tools and end-user clinical s<strong>of</strong>tware developed, area by area. 

Chapter 1 contains a description <strong>of</strong> the anatomy <strong>of</strong> UX and LX, with a focus 

on the main pathologies encountered in the overall project; a description <strong>of</strong> the 

<strong>motion</strong> <strong>analysis</strong> systems <strong>based</strong> on inertial and magnetic sensors-<strong>based</strong> systems. 

Chapter 2 describes the <strong>development</strong> <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on 

stereophotogrammetry and specifically designed for the <strong>analysis</strong> <strong>of</strong> 

compensation strategies in patients with shoulder pathologies, together with a 

test-retest reliability <strong>analysis</strong>. A modified version <strong>of</strong> this protocol is also 

described for the <strong>analysis</strong> <strong>of</strong> the performance and control <strong>of</strong> myoelectric elbow 

prosthesis (Dynamic Arm, Ottobock Healthcare, Germany). Moreover, other 

applications <strong>of</strong> the protocol are presented, such as a preliminary study <strong>of</strong> the 

potential limitations in the use <strong>of</strong> clinical evaluation scales for the <strong>analysis</strong> <strong>of</strong> 

compensation strategies. Finally, the s<strong>of</strong>tware (UpLiFE Toolbox) adopted 

during the <strong>development</strong> and validation <strong>of</strong> the protocol is presented. 

Chapter 3 illustrates the <strong>development</strong> <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on 

stereophotogrammetry designed for the 3D kinematic and kinetic <strong>analysis</strong> on 

transfemoral amputees during walking. Based on the CAST technique [16], the 

protocol was developed considering different types <strong>of</strong> knee prostheses (C-Leg 

reactive knee, Ottobock Healthcare, Germany and Power Knee active knee, 

Ossur, Iceland). Moreover, a comparison between two different methods for the 

estimation <strong>of</strong> joint moments and inertial parameters <strong>of</strong> the knee prostheses is 

28

presented. The s<strong>of</strong>tware (LoLiFE Toolbox) adopted for the experiments is also 

presented. 

In Chapter 4 "Outwalk", a protocol <strong>based</strong> on IMMS specifically designed for 

the functional evaluation <strong>of</strong> cerebral palsy children and lower limb amputees 

during walking, is described. Outwalk was validated on a population <strong>of</strong> 

transtibial amputees with very good results. For transfemoral amputees, when 

magnetic distortions due to ferromagnetic material in the prostheses occurred, a 

new sensor-fusion algorithm was applied and its validation on a subject is 

presented. The method is described in Chapter 6. Finally, Outwalk Manager, 

end-user clinical s<strong>of</strong>tware for the application <strong>of</strong> Outwalk protocol in clinical 

settings is presented. 

Chapter 5 presents a brief description <strong>of</strong> the functional evaluation protocol 

<strong>based</strong> on IMMS, designed for the <strong>analysis</strong> <strong>of</strong> the scapulo-humeral rhythm for 

patients with shoulder pathologies. This chapter is focused on the inter and 

intra-operator reliability <strong>analysis</strong> and the comparison in terms <strong>of</strong> accuracy 

between the IMMS adopted and a 3D kinematics measurement system for 

quasi-static movements <strong>of</strong> the scapula [27]. 

Chapter 6 is entirely focused on the test and validation <strong>of</strong> a novel method for 

the description <strong>of</strong> the 3D kinematics <strong>of</strong> knee and ankle joints when using 

IMMS and magnetic distortions occur. A new Kalman filter-<strong>based</strong> algorithm 

developed by Xsens Technologies B.V. is proposed for the accurate estimation 

<strong>of</strong> 3D joint orientation without magnetometers. Finally, another application <strong>of</strong> 

this method and how it can be interfaced with UX and LX <strong>protocols</strong> described 

in the previous chapters are presented. 

Separate chapter (Chapter 7) provides a description <strong>of</strong> some methodologies 

developed and common among the <strong>protocols</strong>. 

Finally, in Chapter 8 the entire work <strong>of</strong> the thesis is discussed and conclusions 

are drawn. 

A complete list <strong>of</strong> publications and presentation at scientific conferences 

produced along the years <strong>of</strong> the Ph.D. project is given as final part <strong>of</strong> this 

thesis. 

29

1.7 Functional anatomy <strong>of</strong> the upper-extremity 

The following sections present a brief overview about the upper extremity 

anatomy and functionally, focusing on the regions <strong>of</strong> main interest in this 

thesis, that is shoulder girdle, and glenohumeral joint. Most <strong>of</strong> this section is 

taken from [28]. Illustrations are taken from [28] and [29]. 

Figure 6 – 3D representation <strong>of</strong> upper extremity 

Figure 6 roughly describes the main body segments comprising the upper 

extremity (clavicle, scapula, humerus, forearm). There are numerous joints in 

the shoulder complex that must be included in any functional activity <strong>of</strong> the 

upper extremity. All joints must be atomically adequate, well controlled by 

muscular action, and have adequate sensory feedback. 

30

1.7.1 Shoulder girdle 

In Figure 7 a schematization <strong>of</strong> the joints forming the shoulder girdle (complex) 

is presented. 

Figure 7– Joints forming the shoulder girdle: 1) glenohumeral joint, 2) suprahumeral joint, 3) 

acromioclavicular joint, 4) scapulocostal joint, 5) sternoclavicular joint, 6) sternocostal joint, 7) 

costovertebral joint (right limb is shown) 

The shoulder girdle, also called shoulder complex, is formed by 7 joints, 

glenohumeral, suprahumeral, acromioclavicular, scapulocostal, 

sternoclavicular, sternocostal, costovertebral. The glenohumeral joint is also 

clinically named as shoulder joint. 

The clavicle rotates about the manubrium sterni, forming the sternoclavicular 

joint. The scapula is the main structure which supports the arm against the 

thoracic wall. The scapula supports the upper extremity and it is directly 

connected with the glenohumeral joint, and it is supported by ligamentous 

structures between the scapula and the clavicle. 

As represented in Figure 8, clavicle acts as a support at the sternoclavicular 

joint (SC), while the scapula is connected to the clavicle through the 

acromioclavicular joint (AC). Without the presence <strong>of</strong> the claviculoscapular 

trapezium (T) and the conoid (C) ligaments, the scapula would naturally rotate 

around AC. When T and C are subjected to trauma, the superior 

acromioclavicular ligament (SAC) replaces their function. 

31

Figure 8 – Representation <strong>of</strong> the ligaments sustaining the scapula 

Because <strong>of</strong> the clavicle is in a crank formation and there is rotation <strong>of</strong> the 

clavicle at the sternal joint, the scapula can elevate up to 60 degrees (Figure 9). 

Beyond the support in static conditions, the scapula dynamically act in 

coordination with the humerus when the upper extremity performs several 

movements: 

 

 

 

forward flexion and backward extension in the sagittal plane; 

abduction/adduction in the frontal plane 

internal/external rotation in the transverse plane 

When the upper extremity performs these movements, the following rotations 

[ 30 ] <strong>of</strong> the scapula cooperates in coordination with the humerus (Figure 10): 

32

medio/lateral (upward-downward) rotation in the frontal plane 

(Figure 11); 

anterior-posterior tilting in the sagittal plane (Figure 12); 

protraction/retraction in the transverse plane (Figure 13) 

Figure 9 – A scapula range <strong>of</strong> <strong>motion</strong> in the frontal plane, imaging that the clavicle does not rotate; 

B scapula range <strong>of</strong> <strong>motion</strong> in the frontal plane, when the clavicle rotates. 

33

Figure 10 – Movements <strong>of</strong> the shoulder complex in coordination with the arm, on the different 

planes 

Figure 11 – Medio/lateral rotation <strong>of</strong> the scapula on the frontal plane (left limb is shown) 

34

Figure 12 – Anterior/posterior tilting <strong>of</strong> the scapula on the sagittal plane (left limb is shown) 

Figure 13 – Protraction/retraction <strong>of</strong> the scapula on the transverse plane (left limb is shown) 

1.7.2 Glenohumeral joint 

The glenohumeral joint, is clinically termed the ―shoulder joint,‖ as most 

functions require movement or stabilization <strong>of</strong> this joint. The glenohumeral 

35

joint contains many tissues that simultaneously are the tissue sites <strong>of</strong> injury or 

impairment. Figure 14 shows the structure <strong>of</strong> the glenohumeral joint. The 

synovial capsule contains synovial fluid to lubricate all the tissues during 

movement. 

In the static shoulder with the arm dependent, the humerus would, by virtue <strong>of</strong> 

gravity and the weight <strong>of</strong> the upper extremity, dislocate downward out <strong>of</strong> the 

shallow glenoid fossa. The glenohumeral capsule retracts when the arm is 

abducted or forward flexed, further allowing instability <strong>of</strong> the joint during these 

movements. The integrity <strong>of</strong> the capsule to stabilize the glenohumeral joint is 

compounded by the structure <strong>of</strong> the capsule. The glenohumeral capsule is very 

thin and has limited flexibility. It is not strong enough to prevent downward 

subluxation if not assisted by the rotator cuff. The head <strong>of</strong> the humerus is thus 

maintained with stability in the glenoid fossa by the combined action <strong>of</strong> the 

rotator cuff and the capsule. 

Figure 14 – The glenohumeral joint 

The ―rotator cuff,‖ is formed by the conjoined tendon <strong>of</strong> the supraspinous, 

infraspinous and teres major muscles, passes over the humerus and attaches to 

its greater tuberosity (Figure 15). 

36

Figure 15 – The tendons forming the ―rotator cuff‖ 

In the static dependent arm, the supraspinous muscle sustains the head <strong>of</strong> the 

humerus in the glenoid fossa by isometric contraction. 

Figure 16 – Rotator cuff muscles and their lines <strong>of</strong> pull 

Figure 16 illustrates the muscles acting on the humeral head. Supraspinous and 

infraspinous muscles abduct and rotate head <strong>of</strong> humerus. Subscapular muscle 

abducts to lesser degree but also rotates and depresses head <strong>of</strong> humerus. 

The term rotator cuff derives from the fact that, in addition to static support <strong>of</strong> 

the dependent arm, the cuff abducts and forward flexes the arm with 

simultaneous rotation as needed to pass by the acromion and coracoacromial 

ligament. 

37

Glenohumeral movement is a complex action dictated by the anatomical 

structures <strong>of</strong> the articulation. As the humerus begins abduction or flexion, it 

moves to a degree ultimately limited by the acromion or the coracoacromial 

ligament or both. 

With no rotation and no scapular <strong>motion</strong>, 90 degrees <strong>of</strong> abduction is possible 

before the greater tuberosity. 

With the arm internally rotated, the greater tuberosity impinges after only 60 

degrees <strong>of</strong> abduction. 

With external rotation, the greater tuberosity passes behind the coracoacromial 

ligament and the acromial process and is able to abduct and elevate to 

approximately 120 degrees. 

Therefore abduction and overhead elevation <strong>of</strong> the arm requires simultaneous 

external rotation <strong>of</strong> the humerus. 

There are muscles that rotate the humerus other than muscles originating from 

the scapula, namely, the latissimus dorsi and the greater and smaller pectoral 

muscles. The movement <strong>of</strong> the glenohumeral joint is a complex action that 

emphasizes the incongruity <strong>of</strong> that joint. As the arm abducts, or forwardposteriorly 

flexes, the head <strong>of</strong> the humerus glides down and forward and 

backward on the glenoid fossa. This is a muscular action <strong>of</strong> the rotator cuff and 

other glenohumeral muscles, such as the deltoid, latissimus dorsi, and the 

greater and smaller pectoral muscles acting in coordination. From total 

dependency (0 degrees) to overhead elevation (180 degrees), the humerus must 

abduct (forward flexion); then it gradually and simultaneously externally 

rotates to avoid the rotator cuff tendon being impinged on the overhanging 

acromion and coracohumeral ligament, known as the ―painful arc‖ between 60 

and 120 degrees. 

1.7.3 Scapulo-humeral rhythm 

Scapulo-humeral rhythm is not a recent concept. Already in 1944, Inman et al. 

[31] illustrated it as reported in Figure 17. 

38

Figure 17 – Scapulo-humeral rhythm during abduction <strong>of</strong> the humerus (top) and forward flexion 

(down) 

Without further scapular <strong>motion</strong> the humerus can abduct and overhead elevate 

to only 120 degrees when the acromion prevents further <strong>motion</strong>. The scapula 

therefore rotates to remove the acromion from obstruction. A ―rhythm‖ has 

been postulated, depicting the degrees <strong>of</strong> scapular rotation as contrasted to the 

degrees <strong>of</strong> glenohumeral rotation. A ratio <strong>of</strong> 2:1-2 degrees <strong>of</strong> glenohumeral 

rotation to every degree <strong>of</strong> scapular rotation, has been formulated. As the 

scapula must rotate 60 degrees, the clavicle, which attaches to the acromion, 

39

must also rotate 45 degrees (Figures 18-19). 

Figure 18 – A) Vertical alignment <strong>of</strong> scapula (S) and humerus (H) about axis <strong>of</strong> acromioclavicular 

joint (ac). B), As abduction occurs, scapula rotates 30 degrees and humerus rotates 60 degrees, for a 

total <strong>of</strong> 90 degrees <strong>of</strong> arm abduction. C), For further arm overhead elevation (180 degrees), scapula 

rotates 60 degrees, and humerus rotates on glenoid fossa 120 degrees. Ratio is thus 2:1. 

Figure 19 - Clavicular Component <strong>of</strong> Scapulo-humeral Rhythm Third (III, top) phase <strong>of</strong> scapulohumeral 

rhythm. Clavicle has elevated 30 degrees without rotation (top right). Fourth (IV, bottom) 

phase <strong>of</strong> rhythm, in which clavicle has rotated 45 degrees and scapulo-humeral (SH) has elevated to 

180 degrees. SCA indicates scapuloclavicular angle; 30 degrees, rotation <strong>of</strong> scapula (S); ScE, 

scapular elevation; and H, humerus. 

40

During the first 30 degrees <strong>of</strong> abduction, the scapula stabilizes the upper 

extremity. However, once this phase has been reached, the scapula and the 

humerus move at a 2:1 ratio <strong>of</strong> movement; thus, for every 2 degrees <strong>of</strong> humeral 

<strong>motion</strong>, there is 1 degree <strong>of</strong> scapular <strong>motion</strong>. Ultimately, the total arm may 

reach full (180-degree) overhead elevation. 

The 60 degrees <strong>of</strong> scapular rotation on the chest wall is allowed by the 

combined <strong>motion</strong>s <strong>of</strong> the sternoclavicular and the acromioclavicular joints, with 

commensurate rotation at each. The muscles that activate the scapulo-humeral 

rhythm are all the scapular muscles and the combined glenohumeral muscles: 

the rotators and the deltoid. The precise rhythm ratio <strong>of</strong> 2:1 has been 

challenged. 

In Figure 20, the EMG activity <strong>of</strong> the elevators at the shoulder joint is 

represented with respect to the degrees <strong>of</strong> elevation <strong>of</strong> the arm. 

Figure 20 – EMG activity <strong>of</strong> the elevators at the shoulder joint 

41

1.7.4 Scapulo-humeral rhythm measurement through <strong>motion</strong> <strong>analysis</strong> 

Though <strong>motion</strong> <strong>analysis</strong>, two different methods can be applied when the 

scapulo-humeral rhythm is <strong>of</strong> clinical importance. 

A method <strong>based</strong> on stereophotogrammetry, described in Chapter 2, is able to 

measure the coordination between the ―shoulder girdle‖ complex and the 

movements <strong>of</strong> the humerus. 

Another method, described in Chapter 5, allows to monitor the scapulo-humeral 

rhythm by measuring the humerus and scapula degrees <strong>of</strong> freedom illustrated in 

Figure 21, the correlation <strong>of</strong> which is able to describe the scapulo-humeral 

rhythm. 

Figure 21 – Representation <strong>of</strong> the humerus and scapula degrees <strong>of</strong> freedom involved in the 

scapulo-humeral rhythm. FE and AA are the humerus flexion-extension in the sagittal plane and the 

abduction-adduction in the frontal plane respectively. MELA and PRRE are the scapula mediolateral 

rotation and the protraction-retraction respectively 

42

1.8 Shoulder pathologies and compensation strategies 

1.8.1 Impingement 

―Impingement‖, together with rotator cuff tear is the most frequent shoulder 

pathology. One <strong>of</strong> the most common problems that can lead to injuries <strong>of</strong> the 

muscles <strong>of</strong> the rotator cuff is given by the conflict between the humeral head 

and a non deformable structure like the coracoacromial arch. The overall area 

between the bones is gradually worn away, causing the characteristic pain 

symptoms and, in cases <strong>of</strong> tendon rupture, a significant functional impotence. 

Basically, the impingement shows itself as an acute inflammation, with 

shoulder fatigue and pain which normally go back after a rest period. In the 

final stage, there is a partial or complete tendon rupture, more or less 

conditioning the functional aspect, typically in patients over forty. 

1.8.2 Rotator cuff tear 

The requirements for a normal functioning <strong>of</strong> the rotator cuff are integrity, the 

presence <strong>of</strong> strong muscles, a normal capsular laxity, healthy cuff tendons, a 

smooth and normal surface under the coracoacromial arch, a thin synovial 

capsule. The malfunction <strong>of</strong> the various structures is the most common source 

<strong>of</strong> shoulder problems. Rupture <strong>of</strong> the rotator cuff can be partial or total, acute or 

chronic, traumatic or degenerative. The degeneration <strong>of</strong> the cuff, such as 

impingement syndrome, almost always begins with a partial rupture <strong>of</strong> the 

supraspinatus near about 1 cm to the greater tuberosity. This is defined as the 

critical area. It seems that repeated breakages <strong>of</strong> small groups <strong>of</strong> fibers not only 

lead to self limiting acute symptoms but also to the progressive weakening <strong>of</strong> 

the rotator cuff, making it more prepared to other breakage. 

1.8.3 Instability 

The instability commonly occurs after rotator cuff injury, because the ligament 

and muscle structures that surround the shoulder joint are not longer able to 

effectively contain and stabilize the humeral head in the glenoid. This 

progressively causes damage to other stabilizing elements. 

1.8.4 Rehabilitation principles 

43

There are specific principles [32] that must be taken into account before 

undertaking a rehabilitation program for an injury or surgery <strong>of</strong> the shoulder 

and these are: 

The first principle is to emphasize the rehabilitation treatment <strong>of</strong> the entire 

complex <strong>of</strong> the shoulder girdle rather than focus only on the glenohumeral 

joint, to obtain an adequate strength and muscular endurance and proper 

neuromuscular control and proprioception. 

The second principle <strong>of</strong> rehabilitation <strong>of</strong> the shoulder is to obtain a stable 

scapular base on which the humerus can act. The scapula strictly relates to the 

position <strong>of</strong> the humeral head during movement and helps to maintain a tricky 

balance between the glenohumeral joint and the scapula with the trunk. This 

relationship ensures the precision <strong>of</strong> movement <strong>of</strong> the humeral head and keeps 

the correct levers muscles <strong>of</strong> the glenohumeral joint. 

The third principle <strong>of</strong> rehabilitation is a therapeutic approach that recognizes 

the shoulder girdle as an integral part <strong>of</strong> the upper and the kinetic chain <strong>of</strong> the 

trunk. This principle is <strong>based</strong> on the first two. The patterns <strong>of</strong> voluntary 

movements, which are performed in daily life, sports, etc. ..., are remarkably 

complex, involving both the action <strong>of</strong> muscles <strong>of</strong> the upper limb, trunk and leg. 

For this reason, treatment should focus on the whole upper body. 

The fourth principle is to perform exercises on the scapular plane, which is 

generally the safest and most convenient, to ensure proper biomechanical 

alignment for the muscles <strong>of</strong> the rotator cuff and avoid overloading the s<strong>of</strong>t 

tissues. On the contrary, exercises performed on the coronal plane may cause 

impingement and overload <strong>of</strong> the tendons <strong>of</strong> the rotator cuff, while the 

exercises performed on the frontal plane can overload the surgically repaired 

anterior structures, such as, for example, the capsule and the glenohumeral 

ligaments. 

The fifth principle is to use short lever arms to strengthen the muscles <strong>of</strong> the 

shoulder, especially at the beginning <strong>of</strong> the rehabilitation program. The loads 

should be applied with the arms close to the body and elbows flexed, to raise 

the lever arms with the progress <strong>of</strong> rehabilitation. 

The sixth principle concerns the adoption <strong>of</strong> postures in which the 

44

neuromuscular recruitment is the most appropriate to strengthen the muscles <strong>of</strong> 

the shoulder girdle, the patterns <strong>of</strong> isolated movements strengthen the weak 

muscles, and patterns <strong>of</strong> movements combined restore functional activities. 

The seventh principle is to set a program <strong>of</strong> exercises that go from simple to 

complex, reproducing the progressive forces and the loads to which the person 

will be subjected to return. 

The eighth and final principle asserts that the achievement <strong>of</strong> functional 

stability will ensure good functional outcome. A shoulder is functionally stable 

when there is a normal range <strong>of</strong> <strong>motion</strong>, muscle strength adequate, a normal 

neuromuscular control and proprioception. The achievement <strong>of</strong> functional 

stability through surgery and rehabilitation will ensure the proper return to the 

desired level <strong>of</strong> functionality or previous injuries. 

1.8.5 Clinical evaluation scales 

In order to quantitatively evaluate the signs and symptoms <strong>of</strong> the shoulder 

pathology, clinical evaluation scales are the first step performed during 

rehabilitation, for measuring the impairment and the disability. The evaluation 

scales are a useful tool as a support to the choice <strong>of</strong> an adequate therapeutic 

process. The most commonly clinical scales adopted for the shoulder evaluation 

are DASH (Figures 22-23), CONSTANT (Figure 24) and ASES (Figure 25). 

DASH [33] is an evaluation scales <strong>based</strong> on scores provided by the patient 

itself. Therefore the output provided is characterized by subjectivity and it is 

tipically adopted to measure disability rather than impairment. CONSTANT 

scale [34] is adopted for measuring impairment and the output provided is a 

quantitative indication <strong>of</strong> pain, stability, strength, and range <strong>of</strong> <strong>motion</strong> <strong>of</strong> the 

overall complex <strong>of</strong> the shoulder. A modified version <strong>of</strong> the CONSTANT scale 

is typically used [35]. 

45

Figure 22 – DASH evaluation scale (main modules) 

46

Figure 23 – DASH evaluation scale (optional modules) 

47

SCORE LEFT RIGHT 

PAIN (15) none 15 

mild 10 

moderate 5 

severe 0 

ADLs (20) Activity Level full work 4 

full recreation/sport 4 

unaffected sleep 2 

Positioning up to waist 2 

up to xiphoid 4 

up to neck 6 

up top <strong>of</strong> head 8 

above head 10 

RoM (10+10+10+10) Forward Elevation 0-30 0 

31-60 2 

61-90 4 

91-120 6 

121-150 8 

151-180 10 

Lateral Elevation 0-30 0 

31-60 2 

61-90 4 

91-120 6 

121-150 8 

151-180 10 

External behind head, elbow forward 2 

behind head, elbow back 2 

top <strong>of</strong> head, elbow forward 2 

top <strong>of</strong> head, elbow back 2 

full lelevation from top <strong>of</strong> head 2 

Internal Rotation lateral thigh 0 

buttock 2 

lumbosacral junction 4 

waist (L3) 6 

T12 vert 8 

interscapular (T7 vert) 10 

POWER (25) 

(Cybex Dynamometer;90° ab) 

TOTAL SCORE (100) 

Figure 24 – CONSTANT evaluation scale 

48

SCORE SC.pa. T1 SC.pa.T2 SCORE 

PAIN none 5 PATIENT SUPINE: 

slight 4 PASSIVE TOTAL ELEVATION (degree) 

after unusual activity 3 PASSIVE EXT. ROT. 

moderate 2 

marked 1 STRENGTH (0-5) anterior deltoid 

complete disability 0 middle deltoid 

not available 

external rotation 

PATIENT SITTING: 

internal rotation 

ACTIVE TOTAL 

ELEVATION (degree) 

STABILITY 

ANTERIOR 

PASSIVE INT. ROT. less than trochanter 1 (5nrm; 4 apprehension; POSTERIOR 

trochanter 2 3 rare sublux; 2 recurrent INFERIOR 

gluteal 3 1 recurrent disloc; 

sacrum 4 0 fixed disloc.;NA) 

L5 5 

L4 6 

L3 7 FUNCTION 

L2 8 use back pocket 

L1 9 4 normal perineal care 

T12 10 3 mild compromise wash opposite axilia 

T11 11 2 difficulty eat with utensil 

T10 12 1 with aid comb hair 

T9 13 0 unable use hand+arm at sh. level 

T8 14 NA not available carry 10-15 lbs. 

T7 15 dress 

T6 16 sleep <strong>of</strong> affected side 

T5 17 pulling 

T4 18 use hand overhead 

T3 19 throwing 

T2 20 lifting 

do usual work 

ACTIVE EXTERNAL ROTATION (degree) do usual sport 

ACTIVE EXTERNAL ROTATION (90 ° ab) (NA) 

Figure 25 – ASES evaluation scale 

49

1.8.6 Compensation strategies 

The above clinical evaluation scales do not provide an indication <strong>of</strong> ―how‖ the 

movements are performed and whether the improvements in the rehabilitation 

treatment are hiding compensation strategies or not. 

In Figure 26 a typical compensation strategy is shown, during abduction <strong>of</strong> the 

humerus. The picture shows the maximum level <strong>of</strong> abduction the subject is able 

to achieve. As you can see, not only the range <strong>of</strong> <strong>motion</strong> is very limited, but the 

abduction range can be reached only moving the overall shoulder complex. 

Figure 26 – Typical compensation strategies due to rotator cuff tear 

How <strong>motion</strong> <strong>analysis</strong> can represent a valid support to the practitioner in the 

evaluation <strong>of</strong> the shoulder, quantitatively uncovering the compensation 

strategies during the rehabilitation treatment, is shown in Chapter 2 and 5. 

50

1.9 Upper-extremity amputations and prosthetic devices 

The purpose <strong>of</strong> this section is to provide a brief overview about upper limb 

amputations and the use <strong>of</strong> prosthetic devices for restoring the functionality. 

Particular attention is given to unilateral limb loss, shoulder and elbow 

replacement with shoulder and elbow prostheses. Although the upper limb 

<strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> described in this thesis can be extended for sport 

applications, sport prostheses will not be described. 

The upper-extremity amputations represent a restricted ensemble <strong>of</strong> the overall 

amputations, in which the lower-limb amputations are predominant. 

As reported by the National Services Scotland [36] and represented in Figure 

27, the most frequent amputation is the upper-digits amputation followed by the 

trans-humeral and trans-radial amputations. Moreover, it is interesting to notice 

that, as reported in Table 3, most <strong>of</strong> the amputations regard males than females. 

Figure 27 – Levels <strong>of</strong> upper limb amputations 

51

Table 2 – Upper limb amputations divided among gender and age groups 

Table 4 reports a subdivision <strong>of</strong> the amputations with respect to the cause they 

are generated by. Most amputations are caused by trauma, while all the other 

causes, including diabetes and neoplasia are not predominant. 

Upper limb amputation mostly involves the 16-54 years old group, followed by 

the 55-64 years old group. 

Table 4 – Causes <strong>of</strong> upper limb amputations, divided among age groups 

52

From the rehabilitation point <strong>of</strong> view, the above data provide interesting 

indications, meaning that not only transhumeral and transradial amputations 

have to be most frequently treated, but the treatments have to be designed 

considering a large age group with different needs, habits, and 

physical/psychological problems. 

1.9.1 Control <strong>of</strong> limb prostheses 

Particularly for high-level unilateral limb loss, the systems adopted for 

prostheses <strong>of</strong>ten use hybrid control (cable, myoelectric, switches or 

combination <strong>of</strong> these) and hybrid power (electric- and body-powered). The 

control <strong>of</strong> upper limb prostheses can be either manual or automatic. The latter 

is called ―artificial reflex‖. The reflex action <strong>of</strong> upper limb prostheses can 

facilitate its use by the subject and at the same time it can improve the 

performances. For upper limb prostheses, the feedback existing inside <strong>of</strong> the 

control loop is typically visual, incidental (like vibrations or socket forces). 

Artificial control may not always be considered the best option, because the 

user must have high confidence in the control system. Moreover, the likelihood 

<strong>of</strong> a system failure must be as low as possible. Finally, the more the joints 

involved inside <strong>of</strong> the loop, the more the control <strong>of</strong> the prosthesis becomes 

complex. For instance, the control <strong>of</strong> prostheses for transradial amputees is 

basically obtained through the movement <strong>of</strong> the residual forearm and the 

elbow. If the hand is a cosmetic prostheses than all the control is focused on the 

elbow. When the hand becomes a myoelectric hand controlled by the residual 

muscles, the overall control must take into account both the elbow (which is 

directly controlled by the user) and the wrist (which is controlled inside <strong>of</strong> the 

control loop). 

There are several requirements for providing a good prosthetic device. These 

are: 

a) the prosthesis should work with low mental loading or subconscious 

control; 

b) the prosthesis must be user-friendly; 

c) multiple functions must be obtained through a simultaneous and 

coordinated control; 

d) the prosthesis must not produce noisy movements; 

e) the control <strong>of</strong> the prosthesis must not interfere with the remaining 

functional abilities <strong>of</strong> the subject; 

53

f) the prosthesis appearance has to be natural. 

Sometimes, all the above requirements cannot be achieved due to practical use 

<strong>of</strong> the prosthesis. For instance, when supporting weights at a certain elbow 

flexion, the energy consumption can be a critical aspect. In this case, some 

mechanical locking (which <strong>of</strong> course creates conflicts with some <strong>of</strong> the above 

requirements) is preferred. 

Table 5 contains a list <strong>of</strong> the sources <strong>of</strong> body inputs which can be adopted in 

controlling a prosthetic device. 

Table 5 – Sources <strong>of</strong> body inputs for controlling a prosthetic device [ 14 ] 

1.9.2 Myoelectric control 

The control source <strong>of</strong> myoelectric control is a small electrical potential from an 

active muscle. This signal is amplified and processed to activate a controller 

(switch or proportional) which transfers power to an electric motor. The motor 

finally drive the system, according to potential thresholds. A practical way to 

receive the myoelectric signal for long periods and with low invasivity is to use 

electrodes on the surface <strong>of</strong> the body. The power transmitted to the electric 

motor derives from an external source, that is an external battery. The signal 

deriving from the muscles and amplified is only used as a control signal. 

The myoelectric control works only when the subject is generating voluntary 

54

muscle action. 

The first transradial myoelectric system commercially available was developed 

by Ottobock Healthcare (Germany) in collaboration with Viennatone (Austria). 

The number <strong>of</strong> sites in which the myoelectric signal is collected and the way in 

which it is adopted in the control loop, is what differentiate each myoelectric 

prosthetic device to the other. For some devices, the motor is actuated 

proportionally to the myoelectric signal; other systems can reverse the direction 

<strong>of</strong> rotation, others can use more than one myoelectric signals to control the 

actions <strong>of</strong> the motors. 

Although myoelectric prostheses allow the user to directly be involved in the 

control <strong>of</strong> the prosthesis and, by a more or less complex combined action <strong>of</strong> 

myoelectric signals, more than one joint and more than one degree <strong>of</strong> freedom 

can be controlled, there are some drawbacks due to the working principle. 

Amputees sometimes prefer to use a passive or cosmetic prosthesis in their 

daily lives. In fact, not always the patient is able to control the myoelectric 

prosthesis in an intuitive way. Some <strong>of</strong> myoelectric prosthetic arm, for 

example, use signals from agonist-antagonist muscles like the biceps and the 

triceps as control signals. This is necessary for driving the degrees <strong>of</strong> freedom 

<strong>of</strong> each joint and to make the transition from one joint to another one. In this 

case the subject has certainly been provided with a much more functional 

prosthesis than a passive one, but he is forced to use muscles which normally 

are used for different purposes than the one related to the prosthesis. Therefore, 

the individual training for using the prosthesis or the simplification <strong>of</strong> the 

prosthetic working principles can solve this problem. 

Moreover, it is not always easy to understand whether the subject is actually 

correctly controlling the prosthesis during the activities <strong>of</strong> the daily living. The 

results <strong>of</strong> the training should be monitored then. 

1.9.3Prosthetic phases 

Preprosthetic phase/diagnostic assessment 

During the diagnostic assessment, the aspects the rehabilitation team focuses on 

are: 

- the identification and verification <strong>of</strong> sufficient EMG signal recognition for 

myoelectric control, sufficient capture <strong>of</strong> excursion for body-powered 

control, or both for hybrid control; 

55

- the verification <strong>of</strong> how the combined use <strong>of</strong> muscles contraction (for 

example cocontration) for transferring the control from a joint to the other 

one is suitable for the specific subject: a specific therapy training or a 

different control scheme could be required; 

- the verification <strong>of</strong> the maximum range <strong>of</strong> <strong>motion</strong> which can be obtained 

through the use <strong>of</strong> the socket in body-powered control prostheses; 

Postprosthetic phase 

After the diagnostic assessment, the rehabilitation team focuses on: 

- when the prosthesis is delivered, the range <strong>of</strong> <strong>motion</strong> and the controls 

which must be compared with the one during the preprosthetic phase 

- the verification <strong>of</strong> the integration <strong>of</strong> the prosthesis into the patient‘s life 

style 

The objectives are first to identify the real needs <strong>of</strong> subject, including 

advantages and disadvantages <strong>of</strong> using a certain type <strong>of</strong> prosthesis; second, 

from a technical point <strong>of</strong> view, the success <strong>of</strong> the prosthesis is determined by a 

good comfort for the patient, the adaptation <strong>of</strong> the prosthesis to his 

characteristics and his skills in controlling the artificial limb [14]. 

For subjects with shoulder disarticulation, one alternative is to adopt an elbow 

prosthesis which can be myoelectrically or mechanically controlled. In both the 

cases the movements <strong>of</strong> the ―shoulder‖ are very limited. Moreover, in the case 

<strong>of</strong> the mechanically controlled prosthesis, the elbow flexion can only be 

generated with sufficient cable tension using biscapular abduction, being the 

glenohumeral flexion not available as a control source. 

In the case <strong>of</strong> transhumeral amputation, the myoelectric prosthesis can be 

controlled by, for instance, biceps and triceps muscles <strong>of</strong> the residual humerus. 

In the latter case, the shoulder is involved not only in supporting all the 

prosthesis, but also in controlling the scapulo-humeral rhythm. When the 

transhumeral amputee is fitted with a body-powered controlled prosthesis the 

glenohumeral range <strong>of</strong> <strong>motion</strong> are <strong>of</strong>ten insufficient for effective control. This 

is the case, for instance, <strong>of</strong> children and people with narrow shoulders. Again 

56

the scapulo-humeral rhythm might be altered. 

Therefore, the need <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> in monitoring the alterations in the 

scapulo-humeral rhythm becomes significant in the case <strong>of</strong> transhumeral 

amputees. In fact, without any previous problem in the scapulo-humeral 

rhythm, the presence <strong>of</strong> the prosthesis can potentially lead to compensatory 

strategies. 

Motion <strong>analysis</strong> can help to assess the above aspects, to study the optimal 

control <strong>of</strong> the prosthetic device and to monitor the improvements in controlling 

it by the subject. In this scenario, from an economic point <strong>of</strong> view, this has also 

positive implications for reducing the hospitalization time. 

On this direction, Chapter 2 proposes a <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on 

stereophotogrammetry applied to a case study when a subject is fitted with a 

Dynamic Arm myoelectric elbow prosthesis (Ottobock Healthcare, Germany). 

Moreover, in the case <strong>of</strong> upper limb prostheses, it would be preferable to 

monitor the subject in controlling the artificial limb during the activities <strong>of</strong> the 

daily living. Motion <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on stereophotogrammetry are not 

suitable for this purpose. An alternative solution which includes the use <strong>of</strong> 

IMMS, is proposed in Chapter 5. 

1.10 Lower-extremity amputations and prosthetic devices 

The purpose <strong>of</strong> this section is to provide a brief overview about lower limb 

amputations and the use <strong>of</strong> prosthetic devices for restoring the functionality. 

Particular attention is given to unilateral limb loss, knee and ankle replacement 

with knee and ankle prostheses. Although the lower limb <strong>motion</strong> <strong>analysis</strong> 

<strong>protocols</strong> described in this thesis can be extended for sport applications, sport 

prostheses will not be described. 

The lower-extremity amputations represent the larger ensemble among the 

overall amputations. 

57

Figure 28 - – Lower limb levels <strong>of</strong> amputations 

Figure 28 shows the percentages <strong>of</strong> level <strong>of</strong> amputations for the lower limb. 

Most lower limb amputations are unilateral, at transtibial level, followed by 

transfemoral. As in the case <strong>of</strong> upper limb amputations, most <strong>of</strong> amputations 

occur in males, than females (Table 6). However, differently from upper limb, 

the lower limb amputations are almost uniformly distributed among the age 

groups, with only the ―less than 16‖ group apart. 

Table 6 – Lower limb amputations divided among gender and age groups 

58

Table 7 - Causes <strong>of</strong> lower limb amputations, divided among age groups 

Again, differently from upper limb, from Table 7, it can be noticed that the 

main cause <strong>of</strong> transfemoral and transtibial amputations is dysvascularities, in 

particular diabetes and non-diabetic arteriosclerosis, while trauma represents a 

small percentage <strong>of</strong> the numbers <strong>of</strong> amputations. 

From the rehabilitation point <strong>of</strong> view, the above data suggest that the design <strong>of</strong> 

treatments for lower limb amputations must be focused on transtibial and 

transfemoral amputations, considering almost every kind <strong>of</strong> age. At the same 

time, not being the trauma the main cause <strong>of</strong> amputation, the rehabilitation 

should take into account additional elements such as vascular diseases. 

59

1.10.1 Amputee gait 

Differently from upper limb prostheses, control <strong>of</strong> lower limb prostheses has 

not been the main aspect to focus on when designing a prosthetic device. 

Interface loads, suspensions and alignment <strong>of</strong> prosthetic devices become 

important elements due to the way in which the lower limb is used (more 

repetitive and stylized movements than the upper limb). 

As an example, the typical gait deviations for transfemoral amputees [14] are 

guidelines which refer to visual observation <strong>of</strong> the different phases <strong>of</strong> gait. For 

each phase <strong>of</strong> gait the ―normal‖ expected parameters are provided and the 

possible causes to the deviation are indicated. 

Instrumental gait <strong>analysis</strong> have been widely adopted for the <strong>development</strong> and 

testing <strong>of</strong> lower limb prosthetic devices, first <strong>of</strong> all because the visual 

observation is not able to detect all the variations occurring in the amputee gait 

with respect to the normal walking. Secondly, because different solutions are 

available for lower limb prosthesis, including hip, knee and ankle prostheses, 

their combination, resulting in different advantages and disadvantages. 

While in the case <strong>of</strong> the upper limb prosthesis, the controlateral limb has not an 

evident influence on the use <strong>of</strong> the prosthesis, in the case <strong>of</strong> the lower limb 

prosthesis both the prosthetic and the unimpaired limb become important. In 

fact, basically the unimpaired limb plays an important role during walking with 

a prosthetic device, and it is not obvious that the unimpaired limb does not 

change its behavior depending on the kind <strong>of</strong> prosthetic devices adopted. 

Many <strong>of</strong> the parameters indicating the expected behavior, like 

- the knee extension stability, the smooth controlled plantar flexion 

during initial contact; 

- step length; 

- hip and knee flexion during preswing phase; 

and many parameters indicating deviation from the normal behaviour, like 

- lateral trunk bending during midstance; 

- pelvic rise during terminal stance; 

- circumduction and vaulting during initial and midswing. 

can be quantitatively measured through instrumental gait <strong>analysis</strong>, providing at 

the same time objective outcomes about the kinematics and kinetics <strong>of</strong> the 

60

prosthetic limb, the unimpaired limb, including all the joints involved. 

Through instrumental <strong>motion</strong> <strong>analysis</strong>, other aspects can be quantitatively 

measured. 

Both the quality <strong>of</strong> the prosthetic device and the residual limb are important for 

the walking ability <strong>of</strong> a person with lower limb amputation. 

For transfemoral amputees it is relevant to measure the range <strong>of</strong> <strong>motion</strong> <strong>of</strong> the 

residual limb, in terms <strong>of</strong> <strong>motion</strong> in the sagittal and coronal planes, in order to 

detect potential muscle contractures [14] and to establish the initial angular 

alignment <strong>of</strong> the transfemoral prosthetic socket. The correct design <strong>of</strong> the 

socket, in fact, provides biomechanical advantages for the amputee during 

walking. 

Another important aspect <strong>of</strong> the amputee walking is knee stability. In 

transfemoral prostheses this is defined as the ability <strong>of</strong> the prosthetic knee to 

remain extended and support the amputee during stance phase. 

Knee instability can potentially lead to unexpected falls. On the other side, 

when the knee is too stable, that is resistant to flexion, it is difficult for the 

amputee to flex the knee during preswing phase, therefore augmenting the 

energy expenditure. 

In order to improve knee stability two options are available: the first option is 

related to the alignment <strong>of</strong> the knee prosthesis with respect to the lateral weight 

reference line. In this case the more posterior the knee joint is placed with 

respect to the socket-ankle line (Figure 29c) the more stable the knee becomes; 

the second option is to use an ankle-foot prosthesis which is able to reduce the 

induced knee flexion moment in the initial contact with the ground. This setting 

replicates what happens during the normal human loco<strong>motion</strong>, where the 

plantar-flexion amortizes the moment produced at heel strike. 

61

Figure 29 – Variation <strong>of</strong> transfemoral alignment and the effect on knee control. A) voluntary knee 

control alignment; B) alignment stable during knee stance but voluntary control at heel strike is 

required, C) alignment stable throughout stance phase, including heel strike 

Motion <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on stereophotogrammetry, providing both 

kinematics and kinetics outcomes, can support the quantitative evaluation <strong>of</strong> the 

aspects previously described, allowing the comparison <strong>of</strong> different prosthetic 

devices (either <strong>based</strong> on different working principle or provided by different 

suppliers). Moreover, instrumental <strong>analysis</strong> becomes useful when 

compensatory mechanisms in the unimpaired limb occur during walking, as 

reported in [37]. 

An example <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on stereophotogrammetry and 

CAST technique [16], specifically designed for the evaluation <strong>of</strong> prosthetic 

knees is presented in Chapter 3. Among the description <strong>of</strong> the protocol, two 

case studies are reported, regarding one transfemoral (fitted with two different 

prosthetic knees) and one transtibial amputee. 

When the assessment <strong>of</strong> amputees‘ walking is focused on the compensatory 

mechanisms occurring during steady-state walking at different speeds [ 38 ], 

<strong>motion</strong> <strong>analysis</strong> performed inside <strong>of</strong> the laboratory is not suitable for these 

purposes. In this case, it would be preferable to perform gait <strong>analysis</strong> out <strong>of</strong> the 

62

laboratory, allowing the measuring <strong>of</strong> long walking paths, when the condition 

<strong>of</strong> steady state <strong>of</strong> walking is reached. A <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on 

IMMS, specifically designed for lower limb amputees, is proposed in Chapter 

4. Among the description <strong>of</strong> the protocol, its validation study on a population <strong>of</strong> 

transtibial amputee is presented. 

1.10.2 Lower limb knee prostheses 

Control <strong>of</strong> lower limb prostheses became important when microprocessor 

controlled prostheses were developed. In these kind <strong>of</strong> prostheses, the state <strong>of</strong> 

the knee during stance phase is monitored in terms <strong>of</strong> how much the knee is 

bending, in which direction, at which speed and where the centre <strong>of</strong> pressure is, 

being the origin <strong>of</strong> the ground-reaction force. In this case, the stiffness <strong>of</strong> the 

knee can be adjusted in order to prevent falling and allow to execute the gait 

cycle correctly. Focusing on knee prostheses, we can first divide them into two 

main groups: reactive knees and active knees. The main difference between 

reactive and active knees is that in the active knee the electronics controls a DC 

motor which actively flexes and extends the knee, <strong>based</strong> on the signals received 

from sensors positioned both on–board <strong>of</strong> the prosthesis (load cell, gyroscope 

and angulometer) and on the unimpaired limb (foot pressure insole and 

gyroscope). On the contrary, in the reactive knee only the stiffness <strong>of</strong> the knee 

is controlled (body powered), as there are no motors which can actively flex 

and extend the knee. 

Reactive knees 

Reactive knee prostheses can be groups into different classes <strong>based</strong> on their 

biomechanical performance. 

Single-axis knee is the simplest knee available, the least expensive and it does 

not need maintenance. However, it does not provide walking stability without 

being controlled by the subject. This is a big limitation considering the 

population <strong>of</strong> elderly amputees. Normally this prosthesis is adopted for 

children, because <strong>of</strong> the shorter lower leg. 

Stance-control knee uses a friction-brake mechanism for improving stability 

during the stance phase. When the subject applies weight on the prosthetic side, 

the mechanism locks the knee (stance phase), while during the swing phase, 

that is the subject applies weight on the unimpaired limb, the mechanism 

63

unlocks the knee allowing knee flexion. This prosthesis is commonly used 

before the subject is fitted with the designed prosthesis. However, this 

mechanism results in an abnormal gait pattern. 

Polycentric knee (Figure 30) is also called ―four-bar knee‖ due to the fact that 

four bars connect four axes points providing multiple articulations. By this 

particular configuration, the instantaneous center <strong>of</strong> rotation <strong>of</strong> the knee falls 

out <strong>of</strong> the knee joint itself, posteriorly and proximally. As previously discussed, 

this position allows an inherently stable knee during early stance phase. When 

the knee is flexed a few degrees the knee center <strong>of</strong> rotation falls in front <strong>of</strong> the 

lateral weight line. The more the knee flexion the more the center moves 

anteriorly and distally. This behavior provides the knee to be easy to flex in late 

stance phase. 

Figure 30 – Polycentric knee 

Fluid-controlled knee is suitable for subject who are capable to walk at 

variable speeds. The working principle include pneumatic or hydraulic unit 

which controls the knee <strong>motion</strong>. This configuration allows to have a smoother 

and more normal swing phase. While the hydraulic knee performances can be 

affected by temperature, the pneumatic knee can be adopted also in cold areas, 

64

although the pneumatic principle is not suitable for vigorous activities. In 

general, hydraulic knees are more expensive than the pneumatic, and require 

more maintenance. 

Hybrid knee is characterized by the stability <strong>of</strong> a polycentric knee together 

with the cadence swing phase control <strong>of</strong> a hydraulic knee. The knee flexion and 

extension resistance can be set by the prosthetist. 

Microprocessor-controlled knee can be basically considered as a hybrid knee 

in which the knee flexion and extension resistance is automatically adjusted by 

a microprocessor, enabling stance stability and swing-phase control. The result 

is a more normal gait pattern with a more efficient gait. 

C-Leg prosthetic knee (Ottobock Healthcare, Germany) 

The Ottobock C-Leg was first introduced in 1999, as microprocessor-controlled 

knee prosthesis. Swing and stance phase variable resistance is changed thanks 

to the information provided by the sensors positioned within the prosthesis 

pylon (Figure 31). The stance is controlled so that the knee prevents a fall. The 

prosthesis can be customized individually by the prosthetist. The C-Leg is also 

suitable for bilateral amputations but not for activities which require active 

flexion <strong>of</strong> the knee, like going upstairs/downstairs and ripid slopes. 

Figure 31 – C-Leg reactive microprocessor-controlled knee prosthesis (Ottobock Healthcare, 

Germany) 

65

Power Knee prosthetic knee (Ossur, Iceland) 

As reactive knee, the electronics <strong>of</strong> the Power Knee controls a 48V DC motor 

which actively flexes and extends the knee, <strong>based</strong> on the signals received from 

load cells, gyroscopes and angulometer mounted on the prosthesis itself and 

from pressure insoles mounted under the unimpaired foot and a gyroscope 

mounted on the ankle <strong>of</strong> the unimpaired limb (Figure 32). 

Since the working principle is <strong>based</strong> on information coming from the 

unimpaired limb, the Power Knee is not suitable for bilateral amputations. 

Activities like going upstairs/downstairs are supported by this kind <strong>of</strong> 

prosthesis. 

Figure 32 – Power Knee active microprocessor-controlled knee prosthesis (Ossur, Iceland) 

C-Leg and Power Knee do not present differences only for what concerns the 

working principle and the activities that can be performed. As it will be 

described in this thesis, C-Leg and Power Knee has very different weight and 

geometry (according to these features the C-Leg is more similar to a human leg 

than the Power Knee), therefore different inertial parameters which can 

influence the kinetics <strong>of</strong> walking. A comparison between C-Leg and Power 

Knee in terms <strong>of</strong> energy consumption is provided by Cutti et al [39]. 

Figure 33 shows a typical knee kinematics and kinetics report on the sagittal 

66

plane during walking. Inter-joint coordination plots can also represent the 

coordination between different joints, providing a clinical index for monitoring 

improvements in the gait pattern [40]. A typical example is shown in Figure 34 

in which the phases <strong>of</strong> the gait cycle are also indicated. 

Figure 33 - Knee kinematics and kinetics on the sagittal plane during normal walking 

67

Figure 34 – Hip-Knee coordination plot during walking 

Figure 35 – Knee flexion-extension angles during several strides, when comparing the prosthetic 

limb fitted with C-Leg (cyan), Power Knee (red) and the unimpaired limb when the prosthesis 

adopted is C-Leg (black) and Power Knee (green) 

68

Figure 35 shows a comparison between C-Leg and Power Knee knee flexionextension 

during normal walking, as result <strong>of</strong> the application <strong>of</strong> the <strong>motion</strong> 

<strong>analysis</strong> protocol described in Chapter 3 on a transfemoral amputee. While the 

unimpaired limb <strong>of</strong> the subject when fitted with C-Leg (black line) and when 

fitted with the Power Knee (green line) shows a similar pattern, the C-Leg 

(cyan curve), as also confirmed in [41], presents its characteristic flat knee 

flexion during stance phase. The Power Knee (red curve) presents a stance 

phase more similar to the one <strong>of</strong> the unimpaired side, although the range <strong>of</strong> 

<strong>motion</strong> is limited. 

In order to produce results during a steady-state walking out <strong>of</strong> the laboratory, 

<strong>motion</strong> <strong>analysis</strong> on C-Leg was also performed using IMMS for a transfemoral 

amputee (Chapter 4). 

69

1.11 Measurement systems <strong>based</strong> on inertial and magnetic 

sensors 

1.11.1 Working principles 

The working principle <strong>of</strong> inertial sensors can be explained through the human 

vestibular system, located in the inner ear, a real biological system <strong>of</strong> 3D 

inertial sensors. Indeed, this system is able to feel the rotations and linear 

accelerations <strong>of</strong> the head and this allows to maintain the position <strong>of</strong> the eyes in 

the environment. Artificial sensors can replicate the above system through 

MEMS (Micro Electro Mechanical Systems) technology (Figure 36) which 

allows miniaturization <strong>of</strong> sensors, accelerometers and gyroscopes and their 

integration into small portable units, inertial platforms, to track the movement 

<strong>of</strong> body segments <strong>of</strong> interest on which they are positioned. 

Figure 36 – Sensor <strong>based</strong> on MEMS technology 

Gyroscopes provide angular velocity measurements <strong>of</strong> the underlying body and 

this information, if integrated over time, gives angular information from a 

known initial condition. The accelerometers provide linear acceleration 

measurements, including the gravitational component. When the angle between 

the sensor and the vertical direction is known, the gravitational component can 

be subtracted and then by numerical integration the linear velocity and position 

over time theoretically can be obtained. 

The advent <strong>of</strong> inertial and magnetic sensors <strong>based</strong> on MEMS technologies has 

found direct application in <strong>motion</strong> <strong>analysis</strong> and despite <strong>of</strong> the working 

principles differ from the systems <strong>based</strong> on stereophotogrammetry, the direct 

measurement <strong>of</strong> kinematic variables in <strong>motion</strong> <strong>analysis</strong> is also <strong>based</strong> on the 

application <strong>of</strong> the inertial system, considered as a ―marker‖ on the body 

segments. However, different ―inertial platforms‖ (a term deriving from 

Navigation) can be obtained from accelerometers and gyroscopes, due to 

70

different combinations <strong>of</strong> hardware and algorithms for the estimation <strong>of</strong> 

orientation, velocity and position. 

Therefore, inertial platforms can work at different levels <strong>of</strong> abstraction. The 

higher the level, the more the processing applied on the inertial sensors output, 

the more the complexity <strong>of</strong> the system and the accuracy in the estimation <strong>of</strong> the 

kinematic quantities. We will refer for this part to the work described in [42], 

which was followed during the 3-years research project, and the nice review 

provided by Tognetti et al in [43]. 

Level 1 

At this level the inertial platform comprises only accelerometers which can be 

used for estimating the inclination <strong>of</strong> the platform with respect to the vertical 

direction, when the acceleration is negligible if compared to the gravity (quasistatic 

condition). Temperature variations may affect the measurement. No 

information about the rotation around the vertical direction is available without 

specific algorithms. The latter can be used for improving the estimation <strong>of</strong> the 

inclination. 

Level 2 

As in level 1, the inertial platform comprises only accelerometers, but a 

Kalman filter-<strong>based</strong> algorithm [44] is adopted for estimating <strong>of</strong>fset, 

acceleration and the gravitational component during dynamic tasks. The 

assumption is that there are no rotations around the axis <strong>of</strong> movement <strong>of</strong> the 

accelerometer. This solution cannot be adopted for real time applications. 

Level 3 

This level comprises the combination <strong>of</strong> accelerometers and gyroscopes for the 

estimation <strong>of</strong> the inclination and, in general, the orientation <strong>of</strong> the inertial 

platform. Without specific algorithms, this solution can still work in real time 

applications and provides an alternative to the use <strong>of</strong> Kalman filtering. 

However, the uncertainty on the gyroscopes bias is critical due to drift during 

integration <strong>of</strong> the angular velocity. Assumptions on initial and final conditions 

during integration [45] can also be made in order to estimate position for 

cycling movements during short periods <strong>of</strong> time [23]. 

Level 4 

As described in [46] Kalman filtering can be adopted for improving the 

estimation <strong>of</strong> orientation, adopting accelerometers and gyroscopes as in level 3. 

71

Some assumptions on the human body movement are made when the system 

models are created. Basically this solution can greatly improve the estimation 

<strong>of</strong> inclination (level 1), and can improve the estimation <strong>of</strong> orientation (level 3) 

but only for short periods <strong>of</strong> measurement, by correcting the gyroscope bias. 

However, with such a configuration, this solution is still affected by the errors 

in the estimation <strong>of</strong> the rotation around the vertical axis (heading error) as in 

level 3. 

Level 5 

At this level we can find all the potential solutions to correct the estimation <strong>of</strong> 

the heading angle (level 4) in order to obtain a good estimation <strong>of</strong> the 3D 

orientation <strong>of</strong> the inertial platform in space. Basically information about earth 

magnetic field is collected from magnetometers and this information, together 

with acceleration and angular velocities from accelerometers and gyroscopes 

are fused together by a Kalman filter-<strong>based</strong> algorithm which provides and 

estimation <strong>of</strong> the 3D orientation <strong>of</strong> the <strong>motion</strong> tracker with respect to a global 

reference frame defined using information about the gravity and the magnetic 

north (Figure 37). 

Figure 37 – Representation <strong>of</strong> the local coordinate system <strong>of</strong> the <strong>motion</strong> tracker and the global coordinate 

systems defined from information from the same inertial and magnetic sensors 

72

This level is intended to provide only a good estimation <strong>of</strong> 3D orientation in 

space, in absence <strong>of</strong> magnetic distortions. When magnetic distortions are 

present and the estimation <strong>of</strong> 3D position is also required, then this solution is 

not sufficient to reach that goal. 

Level 6 

When the magnetic field is distorted but homogenous, a magnetic field 

mapping procedure can be adopted [47,48]. Moreover, new improvements in 

the algorithm can provide robustness to rapid and strong variations <strong>of</strong> the 

magnetic field. When the magnetic field becomes non homogenous and the 

variation <strong>of</strong> the magnetic field cannot be predicted a priori, other Kalman filter<strong>based</strong> 

algorithms can be adopted, like KiC algorithm (Chapter 6) or the method 

presented by [49]. In particular KiC (Kinematic Coupling) uses assumptions on 

the behaviour <strong>of</strong> a joint during dynamic tasks, in order to provide a good 

estimation <strong>of</strong> 3D joint orientation. 

The above solutions provide only a good estimation <strong>of</strong> 3D orientation in space. 

If the estimation <strong>of</strong> 3D position is required, these solutions are not sufficient to 

reach that goal. 

Level 7 

When the estimation <strong>of</strong> 3D position is also required, a biomechanical model 

can be used within the systems in level 5 or 6, depending on the magnetic field 

conditions. An example <strong>of</strong> 3D <strong>motion</strong> tracking system providing 3D position 

and orientation is described in [50]. In this, both Kalman filter solutions 

described in level 6 can be adopted. 

Level 8 

When the accuracy in the estimation <strong>of</strong> 3D position using the system in level 7 

is not sufficient for some applications, additional <strong>motion</strong> capture systems can 

be combined and new fusing algorithms can be adopted for obtaining a better 

estimate <strong>of</strong> position. An example <strong>of</strong> this approach, adopting an UWB (Ultra 

Wide Band) system is proposed in [12]. 

Although this configuration can contain all the best methods for estimating 

position and orientation in space, the overall system is not completely portable, 

therefore it becomes not ubiquitous. 

Examples <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> systems <strong>based</strong> on IMMS are provided in the next 

sections. 

73

1.11.2 Motion <strong>analysis</strong> systems <strong>based</strong> on IMMS 

By analyzing the different operating levels described in the previous section, 

and from what discussed in section 1.2 <strong>of</strong> this chapter it can be worth to point 

out the following concepts. 

First, although the manual operations the user has to perform when measuring 

using inertial and magnetic platform are less complex than the procedures using 

stereophotogrammetry (e.g. marker labeling), measurements obtained through 

inertial and magnetic platforms may still be affected by errors due to the 

positioning on the body segments, s<strong>of</strong>t tissue artifacts or systematic errors due 

to the technology adopted. 

Second, the knowledge about the human body is, in the case <strong>of</strong> 

stereophotogrammetry, only required when developing a <strong>motion</strong> <strong>analysis</strong> 

protocol starting from information about points trajectories, which is the output 

<strong>of</strong> stereophotogrammetric systems. At most, the specific body segments to be 

studied and the specific activity to be measured will influence the marker set 

adopted and/or the positioning <strong>of</strong> the markers (which are, again, parts <strong>of</strong> the 

protocol itself [13]). In the case <strong>of</strong> systems <strong>based</strong> on inertial sensors, the 

knowledge about the human body and the specific application <strong>of</strong> the system, 

join to the ensemble <strong>of</strong> methods adopted for the estimation <strong>of</strong> the kinematic 

quantities, therefore within the measurement system itself. This is an important 

concept to consider when dealing with new inertial sensors-<strong>based</strong> systems. The 

need to develop specific algorithms/methods and s<strong>of</strong>tware for the use <strong>of</strong> these 

systems - for specific applications – is as much important as the <strong>development</strong> <strong>of</strong> 

<strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on them. 

Third, <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on stereophotogrammetry have the 

advantage that information about specific external or internal landmarks 

through the estimation <strong>of</strong> position, can be used for the construction <strong>of</strong> an 

anatomical coordinate system which can describe the movements <strong>of</strong> the 

underlying bone. Not all the <strong>motion</strong> capture systems <strong>based</strong> on inertial and 

magnetic sensors can provide an accurate estimation <strong>of</strong> the position and 

therefore it is not always possible to relate the position <strong>of</strong> anatomical landmarks 

to the position <strong>of</strong> the inertial <strong>motion</strong> tracker system. Functional methods 

[51,52] are therefore more frequently adopted for the description <strong>of</strong> body 

segment kinematics. The calibration <strong>of</strong> anatomical landmarks with respect to 

74

the <strong>motion</strong> tracker through special although time-consuming procedures [53] is 

possible but not suitable for applications in clinical settings. In general, 

although ―anatomical methods‖, <strong>based</strong> on the definition <strong>of</strong> coordinate systems 

through anatomical landmarks, could, in principle, provide data with high 

comparability (Figure 38), at the same time cross-talk effect in the estimation <strong>of</strong> 

the 3D kinematics may occur [54]. ―Functional methods‖, <strong>based</strong> on the 

definition <strong>of</strong> coordinate systems through the estimation <strong>of</strong> mechanical axes <strong>of</strong> 

rotations, the measurement <strong>of</strong> gravity, or the alignment <strong>of</strong> the <strong>motion</strong> tracker to 

the body segments, could, in principle, provide data with low comparability but 

at the same time the cross-talk effect is minimized. In fact, ―only rotations 

obtained by decomposing the orientation around functional axes can give 

indications <strong>of</strong> the real joint rotations‖ [13], providing a more representative 

kinematics. However, functional methods are <strong>based</strong> on the estimation <strong>of</strong> 

functional axes <strong>of</strong> rotation through the execution <strong>of</strong> a specific task. 

Figure 38 – Comparison <strong>of</strong> functional and anatomical methods for describing 3D kinematics, in 

terms <strong>of</strong> kinematic cross-talking and data comparability 

As this thesis will describe, the above concepts require that the <strong>development</strong> <strong>of</strong> 

<strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on inertial sensors is challenging, but this is 

not in contrast with any other <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on 

stereophotogrammetry, because the latter can be necessary for specific 

applications, and also needful for evaluating the <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> 

<strong>based</strong> on inertial sensors. Again, functional and anatomical descriptions <strong>of</strong> 

kinematics have pros and cons which define different targets and contexts <strong>of</strong> 

application. An example <strong>of</strong> what asserted here is provided in Chapter 4, in 

75

particular for <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> focused on applications on amputees. 

Different inertial platforms can be obtained from the combined use <strong>of</strong> 

accelerometers, gyroscopes and magnetometers, together with different 

specifications <strong>of</strong> hardware and embedded algorithms for the estimation <strong>of</strong> 

orientation, velocity and position. 

We can divide them into three main groups: 

1) ―Inertial Measurement Unit‖ which includes inertial sensors such as 

accelerometers and gyroscopes. 

2) ―Magnetic Measurement Unit‖ which is only <strong>based</strong> on magnetic sensors 

3) ―Inertial and Magnetic Measurement Unit‖ which includes inertial sensors 

and magnetometers 

1.11.2.1 Inertial Measurement unit 

Levels 2, 3 and 4 previously described can be ascribed to inertial measurement 

units. Examples <strong>of</strong> their application are easily available in literature. Basically 

the applications regard the direct use <strong>of</strong> inertial sensors output or the estimation 

<strong>of</strong> 3D orientation <strong>based</strong> on them. De Vries et al [55] presented a review on 

accelerometer-<strong>based</strong> systems adopted for physical activity monitoring. Nyan et 

al [56] explored the possibility to adopt gyroscopes for that purpose. Mokkink 

et al [57] studied the reproducibility and validity <strong>of</strong> the Dynaport system 

(McRoberts, The Netherlands), <strong>based</strong> on accelerometers for application on knee 

osteoarthritis and compared the results with visual observation outcomes [58]. 

Coley et al [59] adopted an inertial measurement unit comprising 

accelerometers and gyroscopes for the estimation <strong>of</strong> shoulder 3D kinematics in 

a population <strong>of</strong> subjects with shoulder pathologies. Zijlstra et al. [60] adopted 

similar solution for the estimation <strong>of</strong> hip abduction moment during walking. 

Giansanti et al [61] developed a <strong>motion</strong> capture system <strong>based</strong> on 

accelerometers and gyroscopes for the <strong>analysis</strong> <strong>of</strong> a limited number <strong>of</strong> dynamic 

tasks. Raggi et al [62,63] developed an algorithm, <strong>based</strong> on gyroscopic signals, 

for the automatic detection <strong>of</strong> gait events during walking, which can be applied 

during walking at different speeds and on different kind <strong>of</strong> patients. Kotiadis et 

al [64] adopted inertial sensing for gait phase detection necessary for 

76

controlling a drop foot stimulator. 

1.11.2.2 Magnetic Measurement unit 

Magnetic measurement units have an old history and despite <strong>of</strong> the obvious 

problems due to magnetic disturbances [65], they still represent a good solution 

for applications in which the estimation <strong>of</strong> position is needed in environments 

in which the use <strong>of</strong> camera-<strong>based</strong> systems is not possible. 

There is not a specific level covering this kind <strong>of</strong> system, being the technology 

adopted deriving only from magnetic field generators and magnetic sensors. 

Some <strong>of</strong> these systems, although limited in the number <strong>of</strong> sensing units 

available and only providing 3D orientation, are completely portable (e.g. 

Minuteman from Polhemus, Vermont, US). 

Other systems (e.g. MotionStar wireless Lite, Ascension Technology, Vermont, 

US, Figure 39) are wireless and provide 3D position and orientation but they 

are not completely wearable by the subject, being the source <strong>of</strong> magnetic field 

required around the working area. 

Figure 39 – MotionStar system from Ascension Technology 

The MotionStar system was adopted by Crosbie et al. [66] for studying the 

scapulo-humeral rhythm <strong>of</strong> a population <strong>of</strong> healthy subjects, showing nice 

results. However, at our knowledge no further studies are available about the 

efficacy <strong>of</strong> the protocol on subjects with shoulder pathologies. Meskers et al 

[67] presented a 3D shoulder kinematics protocol using an electromagnetic 

system for quasi-static measurements. McClure et al [68] examined the 

scapular kinematics using a system from Polhemus during dynamic tasks. Mills 

et al. [69] developed a gait <strong>analysis</strong> protocol <strong>based</strong> on Liberty electromagnetic 

system (Polhemus Vermont, US) (Figure 40). In literature techniques for 

calibration <strong>of</strong> electromagnetic devices are also available [70]. 

77

Figure 40 – Liberty electromagnetic system from Polhemus (US) 

1.11.2.3 Inertial and Magnetic Measurement unit 

Levels from 5 to 7 can be ascribed to inertial and magnetic measurement unit. 

Through specific algorithms, this kind <strong>of</strong> system presents the advantage <strong>of</strong> 

being able to provide an accurate estimation <strong>of</strong> the 3D orientation whilst the 

disadvantage <strong>of</strong> the magnetic measurement unit is the main issue. A specific 

solution to this problem is described in Chapter 6. 

Results showed that the root mean square difference in the orientation 

estimation using this solution with respect to a stereophotogrammetric system 

is 2 degrees [47]. 

The output <strong>of</strong> magnetometers can be affected by magnetic distortions in the 

environment, although several approaches can be adopted to augment 

robustness to the orientation estimation in certain conditions. With no 

compensation for the disturbances, the root mean square difference in the 

estimation <strong>of</strong> the orientation, with respect to a stereophotogrammetric system, 

can reach 10 degrees [47]. 

78

Figure 41 – InertiaCube2+ from Intersense (US) 

An example <strong>of</strong> system <strong>based</strong> on inertial and magnetic measurement units, 

providing 3D orientation and adopted in biomechanics are shown by Foxlin et 

al [71] in which InertiaCube (Intersense, US) system (Figure41) is adopted for 

pedestrian <strong>motion</strong> tracking. Raggi et al [25] studied walking regularity in 

transfemoral amputees, using MTx (Xsens Technologies B.V., The 

Netherlands) (Figure 42). 

Figure 42 – MTx unit from Xsens Technologies B.V. (NL) 

IMMS can be connected together forming a network <strong>of</strong> units which send 

synchronized data via cable or wireless to a data logger communicating with a 

computer. This is the case <strong>of</strong> the Xbus Kit <strong>based</strong> on MTx developed by Xsens 

Technologies B.V. (Figure 43). Noort et al [72] adopted the Xbus kit for 

application on cerebral palsy children and knee osteoarthritis [73]; Schepers et 

al [23,74] developed a system for estimating the center <strong>of</strong> mass during walking 

adopting force sensors and MTx. Picerno et al [53] developed a gait <strong>analysis</strong> 

protocol in which a special calibration procedure is adopted in order to define 

coordinate systems through the estimation <strong>of</strong> anatomical landmarks, although 

adopting a procedure not suitable in clinical settings. Monaghan et al [75] 

79

adopted MT9B (the IMMS produced before the advent <strong>of</strong> MTx) from Xsens 

Technologies B.V. for the functional electrical stimulation <strong>of</strong> the triceps surae 

during gait with interesting results. 

Figure 43 – Xbus kit from Xsens Technologies B.V. (NL) 

Zhou et al [26,76] has proposed the use <strong>of</strong> 3D tracking systems from Xsens 

Technologies as a monitoring system for measuring real-time movements <strong>of</strong> 

upper limb so that this information can be integrated into a home-rehabilitation 

service system in order to assess the outcomes <strong>of</strong> rehabilitation during the 

activities <strong>of</strong> daily living. Of course this kind <strong>of</strong> application highlights the 

problems related to inertial and magnetic measurement units: the duration <strong>of</strong> 

the monitoring produces an increasing in the drift errors. To overcome this 

limitations, the author proposes the introduction <strong>of</strong> additional constraints 

related to the anatomy <strong>of</strong> the subject. 

80

Figure 44 – FAB system from Biosyn (Canada) 

Full body systems are also available, like the Functional Assessment <strong>of</strong> 

Biomechanics (Biosyn, Canada) (Figure 44) or MVN Biomech (Xsens 

Technologies B.V.) (Figure 45). The latter belongs to the group <strong>of</strong> system <strong>of</strong> 

level 7 and moreover, providing accurate estimation <strong>of</strong> position together with 

UWB technology [12], to level 8. 

Figure 45 – MVN Biomech from Xsens Technologies B.V. (NL) 

The accuracy <strong>of</strong> the Xbus kit <strong>based</strong> on MT9B and MTx was assessed and 

<strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> were developed starting from them, as Chapter 4 

(applications on UX) and 5 (applications on LX) will describe. 

81

1.12 References 

1. Biondi E. Introduzione all`Ingegneria Biomedica, Collana di Ingegneria 

Biomedica n. 2. 

2. Saleh M, Murdoch G. In defence <strong>of</strong> gait <strong>analysis</strong>. Observation and 

measurement in gait assessment. J Bone Joint Surg Br. 1985;67(2):237-241. 

3. Cappozzo A. Stereophotogrammetric system for kinesiological studies. 

Medical and Biological Engineering and Computing. 1983;21(2):217-223. 

4. Lorenzo Chiari, Ugo Della Croce, Alberto Leardini, Aurelio Cappozzo. 

Human movement <strong>analysis</strong> using stereophotogrammetry: Part 2: Instrumental 

errors. Gait Posture. 2005;21(2):197-211. 

5. Y Ehara, H Fujimoto, S Miyazaki, S Tanaka, S Yamamoto. Comparison <strong>of</strong> 

the performance <strong>of</strong> 3D camera systems. Gait Posture. 1995;3(3):166-169. 

6. Yoshihiro Ehara, Hiroshi Fujimoto, Shinji Miyazaki, et al. Comparison <strong>of</strong> 

the performance <strong>of</strong> 3D camera systems II. Gait Posture. 1997;5(3):251-255. 

7. Vicon, https:\\www.vicon.com. 

8. Moore GA. Crossing the Chasm: Marketing and Selling High-Tech Products 

to Mainstream Customers. Revised. HarperBusiness; 1999. 

9. Mancini M, Zampieri C, Carlson-Kuhta P, Chiari L, Horak FB. Anticipatory 

postural adjustments prior to step initiation are hypometric in untreated 

Parkinson's disease: an accelerometer-<strong>based</strong> approach. Eur. J. Neurol. 

2009;16(9):1028-1034. 

10. Busser HJ, Ott J, van Lummel RC, Uiterwaal M, Blank R. Ambulatory 

monitoring <strong>of</strong> children's activity. Med Eng Phys. 1997;19(5):440-445. 

11. Xsens Technologies B.V. - https://www.xsens.com. 

12. Hol J, Dijkstra F, Luinge HJ, Schon T. Tightly coupled UWB/IMU pose 

82

estimation. Ultra-Wideband, 2009. ICUWB 2009. IEEE International 

Conference on. 2009:688 - 692. 

13. Kontaxis A, Cutti A, Johnson G, Veeger H. A framework for the definition 

<strong>of</strong> standardized <strong>protocols</strong> for measuring upper-extremity kinematics. Clinical 

Biomechanics. 2009;24(3):246-253. 

14. MD DGS, CPO JWMM, MD JHB. Atlas <strong>of</strong> Amputations and Limb 

Deficiencies. 3rd ed. Amer Academy <strong>of</strong> Orthopaedic; 2004. 

15. Cutti A, Gar<strong>of</strong>alo P, Janssens K, Davalli A, Sacchetti R. Biomechanical 

<strong>analysis</strong> <strong>of</strong> an upper limb amputee and his innovative myoelectric prosthesis: a 

case study concerning the Ottobock ―Dynamic arm‖. Orthopaedie Technik 

(Quarterly). 2007;(1):6-15. 

16. Cappozzo A, Catani F, Croce UD, Leardini A. Position and orientation in 

space <strong>of</strong> bones during movement: anatomical frame definition and 

determination. Clin Biomech (Bristol, Avon). 1995;10(4):171-178. 

17. Fantozzi S, Gar<strong>of</strong>alo P, Cutti A, Stagni R. 3D joint moments in 

transfemoral and transtibial amputees: when is the "ground reaction vector 

technique" an alternative to inverse dynamics Submitted to Gait & Posture. 

18. Gar<strong>of</strong>alo P, Cutti AG, Filippi MV, et al. Inter-operator reliability and 

prediction bands <strong>of</strong> a novel protocol to measure the coordinated movements <strong>of</strong> 

shoulder-girdle and humerus in clinical settings. Med Biol Eng Comput. 

2009;47(5):475-486. 

19. Cutti A, Gar<strong>of</strong>alo P, Parel I, Fiumana G, Porcellini G. Intra- and inter-rater 

reliability <strong>of</strong> the scapulohumeral rhythm measured by a protocol <strong>based</strong> on 

inertial and magnetic sensors. Gait & Posture. 2009;30:S17. 

20. Gar<strong>of</strong>alo P, Raggi M, Ferrari A, Cutti AG, Davalli A. Measure <strong>of</strong> the 3D 

gait kinematics in real-life environments through the Outwalk protocol: 

Development <strong>of</strong> the end-user clinical s<strong>of</strong>tware. Gait & Posture. 2009;30:S132- 

S133. 

21. Winter DA. Biomechanics and Motor Control <strong>of</strong> Human Movement. 2nd ed. 

83

Wiley-Interscience; 1990. 

22. Fantozzi S, Gar<strong>of</strong>alo P, Cutti A, Stagni R, Davalli A. Inverse dynamics vs 

ground reaction force vector methods: Application on lower limb amputees. 

Gait & Posture. 2009;30:S61-S62. 

23. Schepers H, Koopman H, Veltink P. Ambulatory Assessment <strong>of</strong> Ankle and 

Foot Dynamics. Biomedical Engineering, IEEE Transactions on. 

2007;54(5):895-902. 

24. Cutti AG, Giovanardi A, Rocchi L, Davalli A, Sacchetti R. Ambulatory 

measurement <strong>of</strong> shoulder and elbow kinematics through inertial and magnetic 

sensors. Med Biol Eng Comput. 2008;46(2):169-178. 

25. Raggi M, Tura A, Rocchi L, et al. Gait <strong>analysis</strong> through inertial sensors in 

transfemoral amputees: Step and stride regularity. Gait & Posture. 

2006;24(Supplement 1):S17-S18. 

26. Zhou H, Hu H, Tao Y. Inertial measurements <strong>of</strong> upper limb <strong>motion</strong>. 

Medical and Biological Engineering and Computing. 2006;44(6):479-487. 

27. Johnson G, Stuart P, Mitchell S. A method for the measurement <strong>of</strong> threedimensional 

scapular movement. Clinical Biomechanics. 1993;8(5):269-273. 

28. Cailliet R. The Illustrated Guide to Functional Anatomy <strong>of</strong> the 

Musculoskeletal System. 1st ed. American Medical Association Press; 2003. 

29. Copeland S. Interactive Shoulder. Primal Pictures Ltd; 2000. 

30. Wu G, van der Helm FCT, Veeger HEJD, et al. ISB recommendation on 

definitions <strong>of</strong> joint coordinate systems <strong>of</strong> various joints for the reporting <strong>of</strong> 

human joint <strong>motion</strong>--Part II: shoulder, elbow, wrist and hand. J Biomech. 

2005;38(5):981-992. 

31. Inman VT, Saunders JB, Abbott LC. Observations <strong>of</strong> the function <strong>of</strong> the 

shoulder joint. 1944. Clin. Orthop. Relat. Res. 1996;(330):3-12. 

32. Vischi I. Integrazione delle scale di valutazione e dell'analisi del movimento 

84

nella riabilitazione delle lesioni complete della cuffia dei rotatori. Faculty <strong>of</strong> 

Medicine and Surgery, Bachelor Thesis. 2007. 

33. Finch E, Brooks D, Stratford PW, Mayo NE. Physical Rehabilitation 

Outcome Measures. Second Edition. Lippincott Williams & Wilkins; 2002. 

34. Pynsent P. Outcome measure in orthopaedics. Butterworth–Heinemann 

Ltd. 

35. Constant CR, Gerber C, Emery RJ, et al. A review <strong>of</strong> the Constant score: 

Modifications and guidelines for its use. Journal <strong>of</strong> Shoulder and Elbow 

Surgery. 17(2):355-361. 

36. NHS NSS. The amputee statistical database for the United Kingdom 2006- 

2007. 

37. Beyaert C, Grumillier C, Martinet N, Paysant J, André J. Compensatory 

mechanism involving the knee joint <strong>of</strong> the intact limb during gait in unilateral 

below-knee amputees. Gait & Posture. 2008;28(2):278-284. 

38. Silverman AK, Fey NP, Portillo A, et al. Compensatory mechanisms in 

below-knee amputee gait in response to increasing steady-state walking speeds. 

Gait & Posture. 2008;28(4):602-609. 

39. Cutti A, Raggi M, Gar<strong>of</strong>alo P, et al. The effects <strong>of</strong> the ‗Power Knee‘ 

prosthesis on amputees metabolic cost <strong>of</strong> walking and symmetry <strong>of</strong> gait— 

Preliminary results. Gait & Posture. 2008;28:S38. 

40. Janis J. Daly, Karen Sng, Kristen Roenigk, Eric Fredrickson, Mark 

Dohring. Intra-limb coordination deficit in stroke survivors and response to 

treatment. Gait Posture. 2007;25(3):412-418. 

41. Segal AD, Orendurff MS, Klute GK, et al. Kinematic and kinetic 

comparisons <strong>of</strong> transfemoral amputee gait using C-Leg and Mauch SNS 

prosthetic knees. J Rehabil Res Dev. 2006;43(7):857-870. 

42. Fiorentino C. Caratterizzazione di una piattaforma inerziale per tracking 

subacqueo, Bachelor Thesis, Department <strong>of</strong> Electronics, Computer Science and 

85

Systems, University <strong>of</strong> Bologna. 2009. 

43. Gruppo Nazionale di Bioingegneria. Sistemi indossabili intelligenti per la 

salute e la protezione dell'uomo. Bologna: Patron; 2008. 

44. Luinge H, Veltink P. Inclination Measurement <strong>of</strong> Human Movement Using 

a 3-D Accelerometer With Autocalibration. IEEE Trans. Neural Syst. Rehabil. 

Eng. 2004;12(1):112-121. 

45. Veltink PH, Slycke P, Hemssems J, et al. Three dimensional inertial sensing 

<strong>of</strong> foot movements for automatic tuning <strong>of</strong> a two-channel implantable drop-foot 

stimulator. Medical Engineering & Physics. 2003;25(1):21-28. 

46. Luinge HJ, Veltink PH. Measuring orientation <strong>of</strong> human body segments 

using miniature gyroscopes and accelerometers. Med. Biol. Eng. Comput. 

2005;43(2):273-282. 

47. Roetenberg D, Luinge H, Baten CTM, Veltink PH. Compensation <strong>of</strong> 

Magnetic Disturbances Improves Inertial and Magnetic Sensing <strong>of</strong> Human 

Body Segment Orientation. 2005. Available at: http://doc.utwente.nl/53057/ 

[Accessed January 4, 2010]. 

48. Roetenberg D, Baten CTM, Veltink PH. Estimating Body Segment 

Orientation by Applying Inertial and Magnetic Sensing Near Ferromagnetic 

Materials. 2007;15:469-471. 

49. Glen Cooper, Ian Sheret, Louise McMillian, et al. Inertial sensor-<strong>based</strong> 

knee flexion/extension angle estimation. J Biomech. 2009;42(16):2678-2685. 

50. Roetenberg D, Slycke PJ, Veltink PH. Ambulatory position and orientation 

tracking fusing magnetic and inertial sensing. IEEE Trans Biomed Eng. 

2007;54(5):883-890. 

51. Luinge HJ, Veltink PH, Baten CTM. Ambulatory measurement <strong>of</strong> arm 

orientation. J Biomech. 2007;40(1):78-85. 



86


53. Picerno P, Cereatti A, Cappozzo A. Joint kinematics estimate using 

wearable inertial and magnetic sensing modules. Gait & Posture. 

2008;28(4):588-595. 

54. Cutti AG, Gar<strong>of</strong>alo P, Davalli A, Cappello A. How accurate is the 

estimation <strong>of</strong> elbow kinematics using ISB recommended joint coordinate 

systems Gait & Posture. 2006;24:S36-S37. 

55. de Vries SI, Bakker I, Hopman-Rock M, Hirasing RA, van Mechelen W. 

Clinimetric review <strong>of</strong> <strong>motion</strong> sensors in children and adolescents. Journal <strong>of</strong> 

Clinical Epidemiology. 2006;59(7):670-680. 

56. Nyan M, Tay F, Tan A, Seah K. Distinguishing fall activities from normal 

activities by angular rate characteristics and high-speed camera 

characterization. Medical Engineering & Physics. 2006;28(8):842-849. 

57. Mokkink LB, Terwee CB, van Lummel RC, et al. Construct validity <strong>of</strong> the 

DynaPort®KneeTest: a comparison with observations <strong>of</strong> physical therapists. 

Osteoarthritis and Cartilage. 2005;13(8):738-743. 

58. Mokkink LB, Terwee CB, Slikke RMAVD, et al. Reproducibility and 

validity <strong>of</strong> the DynaPort KneeTest. Arthritis Care & Research. 2005;53(3):357- 

363. 

59. Coley B, Jolles BM, Farron A, et al. Outcome evaluation in shoulder 

surgery using 3D kinematics sensors. Gait & Posture. 2007;25(4):523-532. 

60. Zijlstra W, Bisseling R. Estimation <strong>of</strong> hip abduction moment <strong>based</strong> on body 

fixed sensors. Clinical Biomechanics. 2004;19(8):819-827. 

61. Giansanti D, Maccioni G, Macellari V. The Development and Test <strong>of</strong> a 

Device for the Reconstruction <strong>of</strong> 3-D Position and Orientation by Means <strong>of</strong> a 

Kinematic Sensor Assembly With Rate Gyroscopes and Accelerometers. IEEE 

Trans. Biomed. Eng. 2005;52(7):1271-1277. 

62. M. Raggi, A.G. Cutti, S. Lippi, et al. Wearable sensors for the real-time 

87

assessment <strong>of</strong> gait temporal symmetry in above-knee amputees: A validation 

study. Gait Posture. 2008;28:S31-S32. 


assessment <strong>of</strong> gait temporal symmetry in above-knee amputees: The ‗SEAG‘ 

protocol. Gait Posture. 2008;28:S31-S32. 

64. Kotiadis D, Hermens H, Veltink P. Inertial Gait Phase Detection for control 

<strong>of</strong> a drop foot stimulator: Inertial sensing for gait phase detection. Medical 

Engineering & Physics. In Press, Corrected Pro<strong>of</strong>. Available at: 

http://www.sciencedirect.com/science/article/B6T9K-4YC8RKW- 

1/2/20ccdd552c13d42a2ea4cb1145553413 [Accessed February 16, 2010]. 

65. Milne AD, Chess DG, Johnson JA, King GJW. Accuracy <strong>of</strong> an 

electromagnetic tracking device: A study <strong>of</strong> the optimal operating range and 

metal interference. Journal <strong>of</strong> Biomechanics. 1996;29(6):791-793. 

66. Crosbie J, Kilbreath SL, Hollmann L, York S. Scapulohumeral rhythm and 

associated spinal <strong>motion</strong>. Clin Biomech (Bristol, Avon). 2008;23(2):184-192. 

67. Meskers C, Vermeulen H, de Groot J, van Der Helm F, Rozing P. 3D 

shoulder position measurements using a six-degree-<strong>of</strong>-freedom electromagnetic 

tracking device. Clin Biomech (Bristol, Avon). 1998;13(4-5):280-292. 

68. McClure PW, Michener LA, Sennett BJ, Karduna AR. Direct 3- 

dimensional measurement <strong>of</strong> scapular kinematics during dynamic movements 

in vivo. J Shoulder Elbow Surg. 2001;10(3):269-277. 

69. Peter M. Mills, Steven Morrison, David G. Lloyd, Rod S. Barrett. 

Repeatability <strong>of</strong> 3D gait kinematics obtained from an electromagnetic tracking 

system during treadmill loco<strong>motion</strong>. J Biomech. 2007;40(7):1504-1511. 

70. Meskers CG, Fraterman H, van der Helm FC, Vermeulen HM, Rozing PM. 

Calibration <strong>of</strong> the "Flock <strong>of</strong> Birds" electromagnetic tracking device and its 

application in shoulder <strong>motion</strong> studies. J Biomech. 1999;32(6):629-633. 

71. Foxlin E. Pedestrian Tracking with shoe-mounted inertial sensors. IEEE 

Comput Graph Appl. 2005;25(6):38-46. 

88

72. Noort JCVD, Scholtes VA, Harlaar J. Evaluation <strong>of</strong> clinical spasticity 

assessment in Cerebral palsy using inertial sensors. Gait & Posture. 

2009;30(2):138-143. 

73. van den Noort J, van der Esch M, Harlaar J. The effect <strong>of</strong> the instrumented 

force shoe on the gait pattern <strong>of</strong> patients with knee osteoarthritis. Gait & 

Posture. 2009;30(Supplement 2):S68-S69. 

74. Schepers HM, Asseldonk VEHF, Buurke JH, Veltink PH. Ambulatory 

Estimation <strong>of</strong> Center <strong>of</strong> Mass Displacement During Walking. 2009;56:1189- 

1195. 

75. Monaghan C, van Riel W, Veltink P. Control <strong>of</strong> triceps surae stimulation 

<strong>based</strong> on shank orientation using a uniaxial gyroscope during gait. Medical and 

Biological Engineering and Computing. 2009;47(11):1181-1188. 

76. Zhou H, Stone T, Hu H, Harris N. Use <strong>of</strong> multiple wearable inertial sensors 

in upper limb <strong>motion</strong> tracking. Medical Engineering & Physics. 

2008;30(1):123-133. 

89

CHAPTER 2 

FUNCTIONAL EVALUATION OF THE 

UPPER-EXTREMITY THROUGH 

STEREOPHOTOGRAMMETRIC SYSTEMS 

ABSTRACT 

2.1 MOTION ANALYSIS ON NON AMPUTEES 

2.1.1 DEVELOPMENT AND VALIDATION OF A PROTOCOL FOR THE EVALUATION OF THE 

COMPENSATION STRATEGIES IN UPPER-EXTREMITY 

2.1.2 APPLICATION SCENARIOS 

2.1.3 REFERENCES 

2.2 MOTION ANALYSIS ON AMPUTEES 

2.2.1 DEVELOPMENT OF A MOTION ANALYSIS PROTOCOL FOR THE KINEMATICS OF UPPER- 

LIMB MYOELECTRIC PROSTHESES 


2.3 DEVELOPMENT OF THE END-USER CLINICAL SOFTWARE FOR THE UPPER- 

EXTREMITY PROTOCOLS BASED ON STEREOPHOTOGRAMMETRY 

2.3.1 UPLIFE - UPPER LIMB FUNCTIONAL EVALUATION TOOLBOX 

91

ABSTRACT 

The <strong>development</strong> <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on stereophotogrammetry 

and specifically designed for the <strong>analysis</strong> <strong>of</strong> the compensation strategies in 

patients with shoulder pathologies is presented, together with its validation in 

terms <strong>of</strong> test-retest reliability. A modified version <strong>of</strong> the protocol above is also 

described for the <strong>analysis</strong> <strong>of</strong> the performances and the control <strong>of</strong> myoelectric 

elbow prosthesis. 

Some further applications <strong>of</strong> the protocol presented are described together with 

a preliminary study about the potential limitations <strong>of</strong> the clinical evaluation 

scales adopted for the <strong>analysis</strong> <strong>of</strong> the compensation strategies. 

Finally, the s<strong>of</strong>tware tools adopted during the <strong>development</strong> and validation <strong>of</strong> 

the protocol are shown. 

92


2.1.1 Development and validation <strong>of</strong> a protocol for the evaluation <strong>of</strong> 

the compensation strategies in upper-extremity 

INTER-OPERATOR RELIABILITY AND PREDICTION 

BANDS OF A NOVEL PROTOCOL TO MEASURE THE 

COORDINATED MOVEMENTS OF SHOULDER-GIRDLE 

AND HUMERUS IN CLINICAL SETTINGS 

Gar<strong>of</strong>alo P, Cutti AG, Filippi MV, Cavazza S, Ferrari A, Cappello A, Davalli A 

Medical & Biological Engineering & Computing, 2009 May; 47(5):475-86 

Abstract 

A clinical <strong>motion</strong> <strong>analysis</strong> protocol was developed to measure the coordinated 

movements <strong>of</strong> shoulder-girdle and humerus (girdle-humeral rhythm – GD-H-R) 

during humerus flexion-extension (HFE) and ab-adduction (HAA), through an 

optoelectronic system. In particular, the protocol describes the GD-H-R with 2 

angle-angle plots for each movement: girdle elevation-depression and 

protraction-retraction versus humerus flexion-extension for HFE, and versus 

humerus ab-adduction for HAA. Each <strong>of</strong> these plots is further divided in two 

subplots, one for the upward and one for the downward phases <strong>of</strong> the 

movement. 

By involving 11 subjects and 2 operators, we measured the protocol‘s interoperator 

reliability which ranged from very-good to excellent depending on the 

angle-angle plot considered (median values <strong>of</strong> the inter-operator coefficient <strong>of</strong> 

multiple correlation for the angle-angle plots higher than 0.94). 

We then computed the subjects‘ average control patterns, together with 

statistically meaningful prediction bands. ±1SD confidence bands were also 

computed and their width ranged from ±0.5° to ±4.6°. Based on these results 

the protocol resulted more sensitive in the measure <strong>of</strong> the GD-H-R than the 

tripod method for scapulo-humeral rhythm tracking. 

93

1. INTRODUCTION 

As soon as patients surgically treated for shoulder instability or rotator-cuff 

tears begin the active mobilisation, an altered coordination between shouldergirdle 

and humerus rotations can be observed (video, on-line material). In 

particular, an abnormal girdle-thoracic kinematics is generally evident during 

humerus elevations, involving range and/or timing <strong>of</strong> elevation-depression and 

protraction-retraction [1]. One <strong>of</strong> the targets <strong>of</strong> rehabilitation is therefore to 

remove these compensatory movements, recovering a normal coordination 

between shoulder-girdle and humerus, which will be referred herein as girdlehumeral 

rhythm (GD-H-R). The standard clinical rating scales for the 

assessment <strong>of</strong> shoulder impairment, i.e. Constant and ASES [2], record 

valuable information about shoulder overall mobility, pain, power, functionality 

and stability. However, they do not address how a movement is performed, nor 

describe the specific alterations <strong>of</strong> the GD-H-R. Quantitative 3D <strong>motion</strong> 

<strong>analysis</strong> with the extraction <strong>of</strong> focused parameters appears a possible solution 

to overcome these limitations, and especially for monitoring the evolution <strong>of</strong> 

the GD-H-R during the different periods <strong>of</strong> rehabilitation. 

Since the shoulder-girdle is formed by the clavicle and scapula, the measure <strong>of</strong> 

the GD-H-R can be decomposed in the measure <strong>of</strong> the clavicle-humeral and 

scapulo-humeral rhythms. A number <strong>of</strong> <strong>protocols</strong> have been developed for this 

purpose and are available in the literature [3,4,5,6,7,8,9,10,11,12,13,14]. 

However, the measure <strong>of</strong> the scapulo-humeral rhythm is not always possible, 

e.g. due to clinical routine constraints which preclude the use <strong>of</strong> currently 

available tracking systems for the scapula [15, 5]. The need to complete the 

acquisitions <strong>of</strong> 1) both shoulders <strong>of</strong> a subject, 2) either muscular or slim, 3) 

non-invasively, 4) within a time interval comparable to the time required to 

complete a patient‘s anamnesis and a clinical scale (i.e. 30 minutes), 5) with no 

constraint on the maximal humeral elevation admissible, 6) with the measure <strong>of</strong> 

multiple repetitions <strong>of</strong> the activities in order to obtain <strong>motion</strong> cycles truly 

representative <strong>of</strong> the subject‘s kinematics, is a concrete example <strong>of</strong> such a 

combination <strong>of</strong> constraints, taken from the authors‘ clinical routine. 

In these same conditions, however, even though the single clavicle-humeral and 

scapulo-humeral rhythms cannot be measured, at least the measure <strong>of</strong> the 

overall GD-H-R could still be possible, with the non secondary advantage for 

the overall GD-H-R <strong>of</strong> being the most evident and visually detectable clinical 

sign. 

Despite its clinical relevance, the quantitative measure <strong>of</strong> the alterations <strong>of</strong> the 

94

overall GD-H-R has received little attention in the movement <strong>analysis</strong> 

literature: no <strong>protocols</strong> are available for this purpose with the exception <strong>of</strong> a 

preliminary proposal by these same authors [16,17,18]. 

The purposes <strong>of</strong> this work were therefore 1) to complete the definition <strong>of</strong> the 

protocol to measure the overall GD-H-R in routinely clinical settings, and 2) to 

determine important metric properties <strong>of</strong> the protocol for clinical applications. 

Concerning this second purpose, since different operators can potentially 

conduct the measurements on a subject on separate sessions, we verified that 

the protocol is robust to a change in operator by measuring its inter-operator 

reliability on a group <strong>of</strong> control subjects. Moreover, since it is important to 

know if a subject is recovering a ―normal‖ GD-H-R with the rehabilitation, we 

measured the average GD-H-R <strong>of</strong> the controls along with statistically 

meaningful prediction bands. An example <strong>of</strong> clinical application <strong>of</strong> the 

controls‘ average response and prediction bands is finally provided in section 5. 

2. DEVELOPMENT OF THE PROTOCOL 

The protocol was intended to measure the overall GD-H-R in routinely clinical 

settings. For this purpose, we developed the protocol assuming the 6 constraints 

detailed in the introduction plus one more: to include in the protocol only motor 

tasks <strong>of</strong> the Constant and ASES scales, in order to ease the cross <strong>analysis</strong> <strong>of</strong> 

clinical and kinematic data. 

Given these 7 constraints, we developed the protocol by addressing the 

following 8 issues: (1) choice <strong>of</strong> the measurement system, (2) identification <strong>of</strong> 

the segments <strong>of</strong> interest, (3) formulation <strong>of</strong> the reference kinematic model <strong>of</strong> 

the shoulder, (4) definition <strong>of</strong> the segments‘ coordinate systems and angles, (5) 

positioning <strong>of</strong> the system‘s sensors on the body <strong>of</strong> a subject, (6) identification 

<strong>of</strong> the activities the subjects have to execute, (7) formulation <strong>of</strong> the outcome 

measure, and (8) specification <strong>of</strong> the data processing. 

2.1 MEASUREMENT SYSTEM 

Since non-invasive methods were required, the protocol was developed for a 

system able to track in time the position <strong>of</strong> skin-mounted sensors, e.g. 

optoelectronic, electromagnetic or ultrasound. Since the reliability <strong>analysis</strong> <strong>of</strong> 

the protocol presented in section 3 was performed with an optoelectronic 

system [19], hereinafter we will only explicitly refer to this type <strong>of</strong> system. 

95

2.2 SEGMENTS OF INTEREST AND REFERENCE KINEMATIC 

MODEL 

Thorax, shoulder-girdle and humerus were considered as the three segments 

forming the shoulder. In particular, the shoulder-girdle was defined as the 

segment connecting the midpoint between the Incisura Jugularis (IJ) and C7, 

and the centre <strong>of</strong> the Glenohumeral Head (GH). Thorax, shoulder-girdle and 

humerus form an open kinematics chain (fig.1a,b). Since segment kinematics is 

<strong>of</strong> interest, the girdle-thoracic <strong>motion</strong> was modelled with two-degrees <strong>of</strong> 

freedom (DoFs), namely elevation-depression (ED) and protraction-retraction 

(PR). The humero-thoracic <strong>motion</strong> was instead modelled with 3 DoFs: flexionextension 

(FE), ab-adduction (AA) and internal-external (IE) rotation. 

Figure 1 Segments <strong>of</strong> interest with their anatomical landmarks (dots), coordinate systems and 

marker placement. a) Frontal plane; b) Sagittal plane; c) Positioning <strong>of</strong> the clusters <strong>of</strong> markers. 

2.3 DEFINITION OF THE SEGMENTS‟ FRAMES AND ANGLES 

To measure the girdle-thoracic and humero-thoracic angles, we defined 

anatomical frames for thorax, shoulder-girdle and humerus 1) consistently with 

the kinematic model, and 2) <strong>based</strong> on the identification <strong>of</strong> relevant anatomical 

landmarks. 

The ISG standard [20] was followed for the anatomical frames <strong>of</strong> thorax and 

humerus (table 1). 

For the shoulder-girdle no standard is available. Therefore, for the shouldergirdle 

the anatomical frame was defined as follows (table 1): X axis from the 

midpoint <strong>of</strong> IJ and C7 to GH; Z axis perpendicular to X and the Y axis <strong>of</strong> the 

thorax; Y perpendicular to X and Z. 

96

Segment 

Axes Definition 

Thorax Y THX = ( (IJ + C7) / 2 - (PX + T8) / 2 ) / ||(IJ + C7) / 2 - (PX + T8) /2|| 

X THX = Y THX 

(T8 – PX) / || Y THX 

(T8 – PX) || : medio-lateral 

Shoulder- 

Girdle 

Humerus 

Z THX = X THX 

YTHX : antero-posterior 

Origin = IJ 

Y GRD = (Z GRD 

X GRD) / ||||: upward 

X GRD = (GH – (IJ + C7) / 2 ))/ ||(GH – (IJ + C7) / 2 ))||: medio-lateral 

Z GRD= (X GRD 

Y THX) / ||||: antero-posterior 

Origin = IJ 

Y H1 = (GH – E) / ||(GH – E)|| : longitudinal 

Z H1 = Y H1 

(EM – EL) / ||Y H1 

(EM – EL)|| : antero-posterior 

X H1 = (Y H1 

Z H1) / ||||: medio-lateral 

Origin = GH 

E= (EL + EM) / 2 

Table 1 Definition <strong>of</strong> thorax, humerus and shoulder-girdle anatomical frames. 

To measure the ED and PR angles, the relative orientation <strong>of</strong> shoulder-girdle 

and thoracic frames was decomposed with the Euler angles sequence YZ‘X‘‘. It 

is important to notice that with the anatomical frame adopted for the shouldergirdle, 

the third Euler angle <strong>of</strong> the sequence is always mathematically null, 

consistently with the mechanical model assumed. 

To measure the FE, AA and IE angles, the relative orientation <strong>of</strong> the humerus 

and thoracic frames was decomposed with the Euler angles sequence XZ‘Y‘‘ 

for movements in the sagittal plane (obtaining FE, AA, IE), and with the 

sequence ZX‘Y‘‘ for movements in the frontal plane (obtaining AA, FE, IE). 

2.4 SENSORS PLACEMENT AND ANATOMICAL CALIBRATIONS 

To link the thorax frame to the system‘s sensors, 4 markers are directly 

positioned over the anatomical landmarks IJ, Processus Xiphoideus (PX), C7, 

and T8. In case <strong>of</strong> visibility problems, 2 markers are positioned as apart as 

possible on the sternum, while the other 2 are positioned as close as possible to 

C7 and T8. The anatomical landmarks are then calibrated [21, 22, 23] with 

respect to this cluster <strong>of</strong> 4 markers. 

To link the humerus frame to the system‘s sensors, the anatomical landmarks 

Epicondylus Lateralis (EL), Epicondylus Medialis (EM) and GH are calibrated 

97

elative to a clusters <strong>of</strong> 4 markers positioned on the humerus (fig. 1c). 

Specifically, three <strong>of</strong> the four markers were positioned posteriorly, on a CO- 

PLUS (BSN Medical, UK) cuff wrapped around the humerus. The 4th marker 

is positioned at the insertion <strong>of</strong> the deltoid as proposed in [24]. The practice 

suggests that this configuration limits the deformation <strong>of</strong> the humerus cluster 

due to elbow flexion. GH was calibrated with respect to the humerus cluster by 

application <strong>of</strong> the regression equations described in [24, 25] . These regression 

equations require the static calibration <strong>of</strong> the acromion (AC) relative to the 

thorax frame. 

The link <strong>of</strong> the shoulder-girdle frame to the system‘s sensors comes as a 

consequence from the previous steps, since the shoulder-girdle frame is <strong>based</strong> 

on anatomical landmarks (GH, IJ and C7) already tracked through the markers 

on humerus and thorax. 

2.5 ACTIVITIES TO BE MEASURED AND NUMBER OF 

REPETITIONS 

The protocol requires the subject under <strong>analysis</strong> to execute two activities 

common to both the ASES and Constant scales, i.e. humerus flexion-extension 

(HFE – sagittal plane) and ab-adduction (HAA – frontal plane). Before starting 

with the measurements, the subject has to familiarize with the movement. 

When confident, the subject is asked to repeat each movement at least 5 times, 

with an interval <strong>of</strong> relax after each repetition. Five repetitions are assumed to be 

sufficient to gather at least 4 cycles truly representative <strong>of</strong> the subject‘s GD-H- 

R. 

2.6 DATA PROCESSING AND OUTPUT OF THE PROTOCOL 

The output <strong>of</strong> the protocol is a graphical representation <strong>of</strong> the GD-H-R during 

HFE and HAA. Specifically, the GD-H-R <strong>of</strong> HFE is described by 4 angle-angle 

plots, 2 for the upward phase <strong>of</strong> the movement (humerus moving cranially) and 

2 for the downward phase (humerus moving caudally): ED vs FE & PR vs FE - 

upward phase, ED vs FE & PR vs FE – downward phase (e.g. see fig. 4a). 

Similarly, the GD-H-R <strong>of</strong> HAA is described by: ED vs AA & PR vs AA - 

upward phase, ED vs AA & PR vs AA – downward phase (e.g. see fig. 4b). To 

reach this representation <strong>of</strong> the GD-H-R, before being plotted FE, AA, ED and 

PR undergo two macro-steps: (1) a segmentation procedure to identify the 

98

epetitions <strong>of</strong> the movement and distinguish between the upward and 

downward phases, and (2) <strong>of</strong>fset removal from ED and PR patterns. 

The details <strong>of</strong> the segmentation algorithm are reported in paragraph 8.1 <strong>of</strong> this 

thesis. The algorithm to remove the <strong>of</strong>fsets <strong>of</strong> ED and PR starts from the output 

<strong>of</strong> the segmentation algorithm and can be simply summarised in 2 steps. Firstly, 

the values <strong>of</strong> the ED angle in time, ED(t), correspondent to the onsets <strong>of</strong> the 

upward phases are considered and their median value is computed ( ). 

Secondly, 

is subtracted to ED(t). Identical steps apply to PR. 

2.7 OPERATIVE STEPS FOR THE APPLICATION OF THE 

PROTOCOL 

The practical application <strong>of</strong> the protocol on one shoulder <strong>of</strong> a patient requires 

the involvement <strong>of</strong> an operator who completes the following steps: 

1) wraps a CO-PLUS cuff around the humerus <strong>of</strong> the subject; 

2) positions the 8 markers required by the protocol on thorax and humerus; 

3) with the subject standing in upright posture with arms relaxed and the elbow 

flexed 45°, calibrates EL, EM and AC (IJ, PX, C7 and T8 only if required); 

4) shows to the subject the HFE activity and lets the subject familiarize with it; 

5) starts the measurement <strong>of</strong> the first movement, checking for the subject to 

relax between repetitions and for possible anomalies in the movement; 

6) stops the acquisitions after 5 visually correct executions <strong>of</strong> the movements; 

7) repeats steps 4 to 6 for HAA. 

Data extraction is then automatically performed via s<strong>of</strong>tware. 

3. ASSESSMENT OF THE PROTOCOL 

The protocol developed was tested in-vivo in two experiments that were 

executed simultaneously on a group <strong>of</strong> control subjects and with the 

99

involvement <strong>of</strong> 2 operators. The first experiment aimed to assess the interoperator 

reliability <strong>of</strong> the protocol. The second experiment aimed to compute 

the average GD-H-R <strong>of</strong> the controls along with statistically meaningful 

prediction bands. 

3.1 SUBJECTS AND OPERATORS 

Eleven able-bodied subjects (30±3 years-old, 9 male, 2 female) participated in 

the experiments after signing an informed consent. A physical examination 

excluded any pathology in the subjects‘ upper-limbs. The experiments also 

involved two operators O1 and O2 with familiarity in the application <strong>of</strong> the 

protocol. Specifically, O1 and O2 had previously lead 5 measurements with the 

protocol. 

3.2 SET-UP, PROCEDURE AND PREMILINARY DATA PROCESSING 

A common set-up and procedure was used for both experiments. Specifically, 

the GD-H-R <strong>of</strong> a side <strong>of</strong> each subject was measured once by operator O1 and 

O2 through the application <strong>of</strong> the protocol described in section 2, by using a 

Vicon MX 1.3 optoelectronic system (Oxford Metrics, UK) with a sampling 

frequency <strong>of</strong> 100Hz. For each <strong>of</strong> the 2 acquisitions, therefore, each subject 

repeated 5 times the 2 movements HFE and HAA, but only the last 4 repetitions 

were used for the subsequent computations. 

The side acquired for each subject was randomly selected as well as the order 

<strong>of</strong> the operators. 

For each subject the acquisitions by O1 and O2 were from 10 to 30 minutes 

apart. The second operator was not aware <strong>of</strong> the position <strong>of</strong> the markers 

adopted by the first. 

The first part <strong>of</strong> data processing was also common to the two experiments. For 

each subject and angle-angle plot, it was firstly selected the range <strong>of</strong> the angle 

on the X-axis common to both the 4 waveforms obtained by O1 and by O2. 

Then 80 equidistant values were selected in this common range and each <strong>of</strong> the 

8 waveforms was interpolated (using a cubic spline) and re-sampled on the 80 

values. 

3.3 EXPERIMENT 1: INTER-OPERATOR RELIABILITY 

The inter-operator reliability <strong>of</strong> the protocol in measuring the GD-H-R was 

100

quantified by means <strong>of</strong> two different parameters: 1) the Coefficient <strong>of</strong> Multiple 

Correlation (CMC), similarly to [26], and 2) the average inter-operator standard 

deviation (IOSD), as described by [10]. The former was used since in recent 

years the CMC is becoming a standard for the measure <strong>of</strong> the repeatability <strong>of</strong> 

waveforms [27, 26, 28,29,30, 31]. The latter was used instead to enable the 

comparison <strong>of</strong> the results <strong>of</strong> this study with those reported in [10] about 

clavicle-humeral and scapulo-humeral rhythm. 

3.3.1 Measure <strong>of</strong> the inter-operator reliability through CMC 

3.3.1.1 Inter-operator reliability within each subject 

The inter-operator reliability <strong>of</strong> the protocol was firstly quantified separately 

for each subject, for each <strong>of</strong> the 8 angle-angle plots provided by the protocol. 

For what follows, it may be useful to recall that in each plot 8 waveforms are 

reported, 4 measured by O1 and 4 measured by O2. 

In details, for each plot we followed the 2 steps below: 

1) we checked if the subject was highly repeatable in the execution <strong>of</strong> the 

movement, both when acquired by O1 and by O2 (intra-subject repeatability). 

To check for the intra-subject reliability we computed the similarity between 

the 4 waveforms acquired by each operator. The similarity was measured 

through the CMC named by Kadaba et al. within-day CMC [27], which will be 

referred to herein as intra-subject CMC. For the computation <strong>of</strong> the intrasubject 

CMC we applied the technique described in [28]. The intra-subject 

CMC is a single scalar value which combines the similarity for both operators, 

i.e. it tends toward 1 if both the waveforms <strong>of</strong> O1 and <strong>of</strong> O2 are similar, and 

toward 0 otherwise. 

Only if the subject presented an intra-subject CMC higher that 0.95 we 

proceeded with step 2. 

2) we evaluated the inter-operator reliability by comparing the similarity <strong>of</strong> the 

8 waveforms obtained by O1 and O2 considered altogether. In particular, the 

inter-operator reliability was evaluated, similarly to [26], through the second 

CMC proposed in [27], i.e.: 

(1) 

101

where 

k = 1,…, A (A=80): differentiate the 80 samples <strong>of</strong> the angle on the X-axis. 

j = 1,…, R (R=4): differentiate the 4 repetitions <strong>of</strong> each movement, in order <strong>of</strong> 

execution. 

i = 1,…, O (O=2): differentiate the waveforms obtained by operator O1 from 

those <strong>of</strong> operator O2. 

is the Y-axis angle correspondent to the k-th X-axis angle, <strong>of</strong> the j-th 

waveforms, obtained by the i-th operator; 

is the average among all the R * O waveforms <strong>of</strong> the subject at the k-th X- 

axis angle; 

is the grand mean <strong>of</strong> all the waveforms from all the operators. 

This CMC will be referred hereinafter as inter-operator CMC. 

As for the intra-subject CMC, when the waveforms are similar, CMC tends to 

1. If the waveforms are dissimilar, CMC tends to 0. Thus, the inter-operator 

CMC measures the reliability <strong>of</strong> the waveforms by the two operators and is a 

combined measure over 8 repetitions. 

It may be observed that the inter-operator CMC can be lowered by a poor intrasubject 

repeatability, i.e. by the dispersion <strong>of</strong> the waveforms. More specifically, 

the inter-operator CMC can be lowered by 1) the low repeatability <strong>of</strong> the 

subject within each acquisition (intra-subject repeatability), and 2) the 

biological variability <strong>of</strong> the subject between the acquisitions <strong>of</strong> the two 

operators. To compensate for the first cause, through step 1) we considered 

only those subjects with excellent intra-subject repeatability. The effect on the 

inter-operator CMC <strong>of</strong> the (limited) biological variability <strong>of</strong> these subjects was 

ascribed to the inter-operator error. To compensate for the second cause, we 

chose a very limited time interval between the two acquisitions (30 minutes as 

worst case). This excluded large between-days variability. As before, the effect 

on the inter-operator CMC <strong>of</strong> the biological variability between the two 

acquisitions was entirely ascribed to the inter-operator error. 

3.3.1.2 Inter-operator reliability among the subjects 

Once assessed the inter-operator reliability for each plot and subject, statistical 

parameters were computed to describe the inter-operator reliability among the 

subjects <strong>of</strong> the study. 

For each plot, the distribution <strong>of</strong> the inter-operator CMC among the subjects 

102

was firstly checked for normality by visual inspection <strong>of</strong> normality plots. Since 

the inter-operator CMC did not generally show a normal distribution among the 

subjects (ceiling effect), the median and interquartile distance (IQD) was 

computed for the subjects and a box & whiskers plot was created. For each 

plot, the protocol inter-operator reliability was then interpreted as follows, 

<strong>based</strong> on the median and IQD range <strong>of</strong> the inter-operator CMC and <strong>based</strong> on 

previous publications [26,27,31]: 

 

 

 

 

0.65 < CMC < 0.75 moderate 

0.75 < CMC < 0.85 good 

0.85 < CMC < 0.95 very good 

0.95 < CMC < 1 excellent 

Finally, to access if certain plots were more reliable than others, we compared 

the inter-operator CMCs <strong>of</strong> the 4 angle-angle plots <strong>of</strong> HFE and <strong>of</strong> the 4 angleangle 

plots <strong>of</strong> HAA through repeated measures non-parametric ANOVA 

(Friedman‘s test). 

3.3.2 Measure <strong>of</strong> the inter-operator reliability through IOSD 

In [10], the inter-operator reliability was measured by computing for each 

subject the numerator <strong>of</strong> the ratio <strong>of</strong> Eq. 1, and this quantity was called interoperator 

variance. The average <strong>of</strong> the inter-operator variance was then 

computed over the subjects. The root square <strong>of</strong> this average, i.e. the average 

inter-operator standard deviation (IOSD), was finally reported. For the seek <strong>of</strong> 

comparison, this same procedure was followed here. 

103

3.4. EXPERIMENT 2: PREDICTION BANDS AND ±1SD CONFIDENCE 

BANDS 

It is <strong>of</strong> clinical interest to know if the angle-angle patterns describing the GD- 

H-R <strong>of</strong> a subject are ―normal‖ patterns. For instance, it can be <strong>of</strong> interest to 

know if a single upward phase <strong>of</strong> ED vs FE is ―normal‖. A ―normal‖ pattern 

will be assumed herein as a pattern which remains within the protocol‘s 

minimal detectable difference band (also called herein ―prediction band‖ or 

MDDB) from the control subjects‘ average pattern. The minimal detectable 

difference [32] band for a given angle-angle plot is a (1-α)% confidence band 

around the average pattern <strong>of</strong> the control subjects, with a clear statistical 

meaning: when a pattern <strong>of</strong> a new subject is outside <strong>of</strong> the prediction band, 

there the subject‘s pattern is different from the control average with a (1-α) % 

probability. MDDB bands are alternative to the more common confidence 

bands used in <strong>motion</strong> <strong>analysis</strong> <strong>based</strong> on ±1SD <strong>of</strong> the population [33], with the 

remarkable advantage for the former <strong>of</strong> being meaningful from the inferential 

statistics viewpoint. As detailed below (sect. 3.4.1), MDDBs are directly related 

to the Standard Errors <strong>of</strong> Measurement (SEM) <strong>of</strong> the protocol [32]. 

For each <strong>of</strong> the 8 angle-angle plots describing the GD-H-R, the control 

subjects‘ average and MDDB was computed as described in section 3.4.1, 

starting from the kinematic data collected by both operator O1 and O2, 

altogether. 

To compare the results from this study with previous works concerning the 

clavicle and scapula-humeral rhythm, we also computed for each plot the ±1SD 

confidence bands as described in [33]. 

3.4.1 Computation <strong>of</strong> an average pattern and its prediction band (MDDB) 

For the description below, it is worth recalling that each angle-angle plot has 80 

values in abscissa (see section 3.2). Since 11 subjects were measured by 2 

operators, and for each subject 4 repetitions were considered, 88 ordinate 

values exist for each abscissa. 

To compute the average pattern and the MDDB for each angle-angle plot, each 

<strong>of</strong> the 80 angle values (<strong>of</strong> FE for HFE and AA for HAA) in abscissa was 

separately considered. 

For the i-th abscissa, the average ordinate from its 88 ordinates was firstly 

computed, and named Mi. 

For the i-th abscissa the MDDB‘s upper and lower values were then computed 

104

from its 88 ordinates as follows: 

a) a two-factors repeated measures ANOVA was executed, with the 

―ordinate angle‖ as dependent variable and with ―operator‖ and ―repetition‖ as 

the two independent variables, with 2 and 4 levels respectively (fig. 2). The 

repeated measures ANOVA method allows to isolate the contribution due to the 

between-subjects variability and to the within-subject variability (MSw). The 

within-subjects variability takes into account all the systematic and random 

errors associated with the application <strong>of</strong> the protocol, i.e. the variability <strong>of</strong> the 

data from repetition-to-repetition, operator-to-operator, a possible interaction <strong>of</strong> 

these two factors, and the total residual. 

Figure 2 - Example <strong>of</strong> data collection for the two-way repeated measures ANOVA performed to 

compute the prediction bands. a) ED angles at 100 degrees <strong>of</strong> FE during 4 repetitions (trials) <strong>of</strong> 

HFE, for each subject; b) data table for ANOVA <strong>analysis</strong>. Operator and Repetition are the two 

factors with 2 and 4 levels, respectively. 

b) the square root <strong>of</strong> MSw was computed, thus obtaining the SEM <strong>of</strong> the 

protocol [34,35,]33 ]: 

(2) 

The SEM is usually referred to as the ―typical error‖ and, being <strong>based</strong> on the 

MSw only, is a fixed characteristic <strong>of</strong> any measure, regardless <strong>of</strong> the sample <strong>of</strong> 

subjects under investigation [35] . 

c) The upper and lower values <strong>of</strong> MDDB are computed from the SEM as 

follows [32] : 

(3) 

105

(4) 

where 

(5) 

is defined as the minimal difference (MD) detectable through the protocol. 

Since MSw incorporated the measure <strong>of</strong> the repetition-to-repetition variability, 

operator-to-operator variability, their interaction and the residual, it can be 

stated that: a new single ordinate value for the i-th abscissa 1) obtained from a 

new subject, and 2) measured by a new operator, is different with a 95% 

probability from the controls average if it falls above the value indicated by 

or below the value indicated by . 

4. RESULTS 

4.1 RESULTS FOR EXPERIMENT 1 – Inter-operator reliability 

The intra-subject CMC was higher than 0.95 for all subjects and angle-angle 

plots, with the exception <strong>of</strong> only two cases, i.e. the intra-subject variability was 

higher than 0.95 in 86/88 cases. Subject 3 presented an intra-subject variability 

<strong>of</strong> 0.89 in the upward phase <strong>of</strong> PR vs FE and subject 4 presented an intrasubject 

variability <strong>of</strong> 0.92 in the downward phase <strong>of</strong> ED vs FE. These two cases 

were excluded from the computation <strong>of</strong> the inter-operator CMC. 

Box & whisker plots with notches <strong>of</strong> the inter-operator CMCs <strong>of</strong> each angleangle 

plot are reported in fig. 3a. Numeric values for the medians and IQDs 

illustrated in fig. 3a are reported in fig. 3b. 

The median values in the boxes are not generally centered and this suggests the 

non normality <strong>of</strong> the distributions, also confirmed by the inspections <strong>of</strong> the 

normality plots. CMCs median values among the 8 angle-angle plots were all 

above 0.94. The lowest 1st quartile was 0.89. These results suggest that the 

inter-operator reliability varied, depending on the angle-angle plot, from ―very 

good‖ to ―excellent‖. 

Friedman‘s tests confirmed that no statistically significant differences exist 

neither between the inter-operator CMCs <strong>of</strong> the angle-angle plots associated to 

HFE (p=0.07), nor to those <strong>of</strong> HAA (p=0.52). 

The IOSD values for the 8 angle-angles plots are reported in table 2. 

106

HFE 

HAA 

Upward Downward Upward Downward 

EDvsFE PRvsFE EDvsFE PRvsFE EDvsFE PRvsFE EDvsFE PRvsFE 

IOSD 

[deg] 

1.78 2.00 2.14 2.28 1.69 1.72 1.74 1.63 

Table 2 - IOSD values describing the inter-operator reliability for each movement (HFE, HAA), 

phase <strong>of</strong> the movement (upward and downward) and angle-angle plot. 

Figure 3 - Inter-operator CMCs among the control subjects for the different angle-angle plots 

describing the GD-H-R. a) box & whiskers plots with notches for all the 8 angle-angle plots coming 

from the protocol; b) Median , quartile values and IQD for the distributions reported in a). 

4.2 RESULTS FOR EXPERIMENT 2 

The controls subjects‘ average pattern, MDDB and ±1SD bands for each <strong>of</strong> the 

107

8 angle-angle plots are reported in fig. 4. To ease the application <strong>of</strong> the protocol 

by other research groups, the numeric data required to plot the average plots, 

MDDBs and ±1SD confidence bands are provided as on-line material 

(Micros<strong>of</strong>t Excel file). MDDBs width ranged among the angle-angle plots 

between ±1.5° to ±7.9°. The SEM ranged therefore between ±0.6° to ±2.8°. The 

±1SD bands ranged instead between ±0.5° to ±4.6°. 

Figure 4 - Control subjects‘ average patterns, MDDBs (prediction bands) and ±1SD confidence 

bands for the 8 angle-angle plots describing the GD-H-R. a) the 4 angle-angle plots associated to 

HFE; b) the 4 angle-angle plots associated to HAA; 

108

5. CLINICAL APPLICATION 

To illustrate an example <strong>of</strong> clinical application <strong>of</strong> the protocol, average patterns 

and MDDBs, let us consider the case <strong>of</strong> a post-surgical patient treated for 

rotator cuff tears. The patient was acquired with the protocol 3 times, i.e. after 

42, 70 and 122 days from the surgery. 

For this patient, the specific clinical questions were: 1) if the ED vs FE pattern 

measured during the 1st, 2nd and 3rd acquisition could be assimilated to a 

normal pattern, and 2) if the rehabilitation was effective in restoring a normal 

ED vs FE pattern. 

Since the intra-subject repeatability <strong>of</strong> the patient for the ED vs FE plot was 

higher than 0.95, the repetition-to-repetition variability is the same <strong>of</strong> the 

population from which the MDDB was computed. In all three acquisitions, 

therefore, the <strong>analysis</strong> could be performed comparing just 1 repetition <strong>of</strong> the 

movement to the average pattern and MDDB. However, to further decrease the 

probability <strong>of</strong> false positives, all 4 repetitions from each acquisition were 

compared to the average controls‘ pattern. 

The ED vs FE patterns for all the three acquisitions are reported in fig. 5. 

Figure 5 - ED vs FE patterns <strong>of</strong> three acquisitions <strong>of</strong> a typical patient recovering from surgery for 

rotator cuff tear, during the HFE task. Controls‘ average and 95% MDDB are provided for 

statistical comparison. 

To answer to the first clinical question, it should be noticed that the patterns <strong>of</strong> 

109

the 1st and 2nd acquisitions felt above the MDDB almost immediately, i.e. 

from very small values <strong>of</strong> FE. This indicates a 95% probability <strong>of</strong> differences 

between the patients and the controls‘ average for almost the entire pattern. In 

the 3rd acquisitions, instead, the patient‘s pattern remained within the MDDB 

until 30° <strong>of</strong> FE. This indicated that only from 30° to the end <strong>of</strong> the range <strong>of</strong> 

<strong>motion</strong> for FE the patient differed from the control‘s average with a probability 

<strong>of</strong> 95%. 

For what concerns the second questions, <strong>based</strong> on 1) the previous 

considerations, 2) the fact that the shape <strong>of</strong> the pattern in the 3rd acquisition is 

closer to the control‘s average shape, and 3) the increase in the range <strong>of</strong> <strong>motion</strong> 

<strong>of</strong> FE from the 1st to the 3rd acquisition, it can be stated that the rehabilitation 

is having a positive influence on the patients but the recovered rhythm <strong>of</strong> FE 

and ED still remains statistically different from the normal pattern. 

6. DISCUSSION AND CONCLUSION 

A <strong>motion</strong> <strong>analysis</strong> protocol was developed to measure the overall GD-H-R in 

clinical settings. Specifically, the protocol allows to measure the coordination 

between humerus FE and shoulder-girdle ED and PR during HFE, and the 

coordination between humerus AA and shoulder-girdle ED and PR during 

HAA. For a single arm the protocol requires only 8 markers, the calibration 

from 3 to 7 anatomical landmarks, and the dynamic acquisition <strong>of</strong> 2 motor 

activities. No specialized equipment is required (e.g. a scapula locator). The 

girdle segment is not <strong>based</strong> on a scapula-tracker over the acromion, but only on 

the calibration <strong>of</strong> GH relative to the humerus cluster. Therefore, the validity <strong>of</strong> 

the measure is not a-priori limited to 120° <strong>of</strong> humerus elevation [5,15]. Overall, 

the protocol fulfils the clinical constraints declared in the introduction. 

To the authors‘ knowledge, the protocol is original in the aims, in the 

description provided <strong>of</strong> the GD-H-R and in the definition <strong>of</strong> the shoulder-girdle 

coordinate system. This last represents an update <strong>of</strong> a previous proposal by 

these same authors [17,23]. With the new coordinate system, the third Euler 

angle <strong>of</strong> the sequence YZ‘X‘‘ used to compute the orientation <strong>of</strong> the shouldergirdle 

relative to the thorax is always mathematically null. The Euler angles 

provided are therefore consistent with the mechanical model assumed for the 

‗joint‘ connecting the shoulder-girdle with the thorax. This was not the case 

with the previous proposal. 

The protocol includes an <strong>of</strong>fset removal step for the ED and PR patterns. The 

same <strong>of</strong>fset is removed from all the 4 repetitions <strong>of</strong> ED (PR). This step was 

110

included since this common <strong>of</strong>fset among the repetitions is strictly dependent 

on the specific anatomy and static posture <strong>of</strong> the subject considered, i.e. by the 

subject-specific position <strong>of</strong> the anatomical landmarks IJ, C7 and GH. Similarly 

to gait <strong>protocols</strong> [27], the main clinical interest was here to detect differences in 

the patterns describing the GD-H-R rhythm. These merely anatomical 

differences between subjects, if not compensated for, would have therefore 

increased the width <strong>of</strong> the MDDB and therefore decreased the sensitivity <strong>of</strong> the 

protocol in detecting differences in the angles patterns. 

Since a common <strong>of</strong>fset was removed from all the repetitions, it should be 

noticed that the variation <strong>of</strong> the initial value <strong>of</strong> ED and PR between the 

different repetitions (which can be <strong>of</strong> clinical interest and affects the value <strong>of</strong> 

the intra-subject variability), has not been removed and is therefore taken into 

account in the MDDBs. 

For what concerns the in-vivo assessment <strong>of</strong> the protocol, the results from the 

first experiment confirmed that the protocol has an inter-operator reliability 

ranging from ―very-good‖ to ‖excellent‖, with no differences between the 

angle-angle plots considered for HFE and HAA. Unfortunately, at the moment 

no other studies are available in upper-extremity <strong>motion</strong> <strong>analysis</strong> which have 

assessed the inter-operator reliability by means <strong>of</strong> the inter-operator CMC. This 

is because the CMC and the inter-operator CMC is currently becoming a well 

accepted standard for this measurement. However, the inter-operator reliability 

results for the IOSD can be compared with previous results by Meskers and coworkers 

[10], and in particular with the IOSD they reported for the claviclehumeral 

and scapula-humeral rhythm. Meskers reported for these two rhythms 

IOSD values ranging from 2.2° to 5.6°, with a mean value <strong>of</strong> 3.4°. In the 

present study, IOSDs ranged from 1.6° to 2.3°, with a mean value <strong>of</strong> 1.9°. This 

demonstrates that the measure <strong>of</strong> the GD-H-R with the protocol presented here 

is more reliable than the measure <strong>of</strong> the scapulo-humeral and clavicle-humeral 

rhythm through the palpation method. A possible explanation <strong>of</strong> these results 

can lay in the fact that the protocol presented here does not require the 

intervention <strong>of</strong> an operator during the acquisitions, but only to set-up the 

procedure. 

Results from the second experiment showed that the average patterns and 

MDDBs generally differed from the upward and the downward phase <strong>of</strong> each 

movement. This confirmed the need to consider the two phases separately in 

the <strong>analysis</strong> <strong>of</strong> the kinematics <strong>of</strong> a subject, as well as the need to always 

incorporate in an upper-extremity protocol a segmentation algorithm. The 

differences between the average patterns in the upward and downward phases 

111

are consistent with previous findings for the scapula-humeral and claviclehumeral 

rhythm [36, 37, 6]. 

Unfortunately, no other studies in upper-extremity <strong>motion</strong> <strong>analysis</strong> have ever 

reported any sort <strong>of</strong> prediction bands. The comparison with previous literature 

has therefore to be <strong>based</strong> on the comparison <strong>of</strong> the ±1SD confidence bands. In 

particular, the best candidate for comparisons appears the paper by Meskers 

and co-workers [15], who have reported ±1SD confidence bands for the 

scapula-humeral rhythm measured with the tripod method, not considering just 

1 observer, but 3, similarly to the present study. Meskers reported the narrowest 

band for the scapula medio-lateral rotation, with 1SD ranging from about 5° to 

10°. In the present study, the widest band presents a 1SD = 4.6° (PR vs AA - 

upward phase). It may be concluded therefore that the protocol presented here 

is more sensitive in the measure <strong>of</strong> the GD-H-R than the tripod method in 

measuring the scapula-humeral rhythm. 

Given the typical alteration <strong>of</strong> the GD-H-R <strong>of</strong> patients recovering from surgery 

for shoulder-instability and rotator cuff tears (see section 5), the widths <strong>of</strong> the 

MDDBs appear adequate to draw solid clinical conclusions, i.e. the protocol 

appears sensitive enough for the application. Further clinical experimentations 

are however required to draw definitive conclusions. 

Remarkably, the MDDBs are generally wider than the confidence bands <strong>based</strong> 

on the SD <strong>of</strong> the population. This suggests that ±1SD confidence bands tend to 

underestimate the uncertainty <strong>of</strong> the average pattern and can lead to an increase 

<strong>of</strong> false positives. This is consistent with previous findings in gait <strong>analysis</strong> [38] . 

In computing the MDDBs, all the repetitions <strong>of</strong> all the subjects were 

considered, as well as the acquisitions <strong>of</strong> each subject by both operators. 

Usually the inter-operator variability is excluded in the computation <strong>of</strong> the 

confidence bands in <strong>motion</strong> <strong>analysis</strong>, excluding therefore an important source 

<strong>of</strong> potential errors. 

The MDDBs computed allow to draw conclusion on a subject <strong>based</strong> on a new 

single measure taken with an operator with comparable experience to those 

who participated in these experiments. Since the MDDBs are <strong>based</strong> on subjects 

with a repetition-to-repetition repeatability <strong>of</strong> at least 0.95, they should be used 

to draw conclusion on a patient <strong>based</strong> on a single measure only if the patient 

presents a similar intra-subject variability (see the section 3.3.1.1). For those 

patients with less repetition-to-repetition repeatability, a simple solution is to 

compare all the repetitions <strong>of</strong> the movement with the controls‘ average and 

MDDB. This increases the probability <strong>of</strong> no difference between the patient‘s 

112

and average controls‘ pattern, thus reducing the probability <strong>of</strong> false positive. 

The definition <strong>of</strong> confidence bands <strong>based</strong> on the Minimal Detectable Difference 

is original but it is <strong>based</strong> on well known statistical methods and procedures 

[32]. The problem <strong>of</strong> estimating confidence bands with a clear inferential 

statistics meaning is receiving increasing interest in the literature [39,40]. In 

particular, Schwartz and co-workers stressed the improper use <strong>of</strong> confidence 

bands <strong>based</strong> on ±1SD for performing statistical tests on the data. The approach 

followed here, is similar to the method proposed by [39], with the advantage <strong>of</strong> 

1) using a standard and well documented 2-way <strong>analysis</strong> <strong>of</strong> variance with 

repeated measures, and 2) defining prediction bands which allows immediate 

graphical statistical tests. Compared to the bootstrap techniques, MDDBs have 

the advantage <strong>of</strong> explicitly excluding the between-subject variability from the 

computation <strong>of</strong> the prediction bands and to allow an angle-by-angle <strong>analysis</strong> <strong>of</strong> 

statistical difference. Moreover MDDBs are <strong>of</strong> very simple implementation. 

113

2.1.2 Application scenarios 

2.1.2.1 Limitations <strong>of</strong> the constant scale in the assessment <strong>of</strong> shoulder 

compensatory strategies 

2.1.2.2 The Evolution <strong>of</strong> Compensation Strategies in Two Patients with 

Shoulder Instability: a Comparative Study through Quantitative Motion 

Analysis 

2.1.2.3 The relation between scapular and shoulder-girdle <strong>motion</strong> in subjects 

with and without shoulder impingement 

114

2.1.2.1 

LIMITATIONS OF THE CONSTANT SCALE IN THE 

ASSESSMENT OF SHOULDER COMPENSATORY 

STRATEGIES 

Gar<strong>of</strong>alo P, Cutti AG, Filippi MV, Cavazza S, Davalli A, Cappello A 

Gait & Posture, vol. 28, suppl. 1, August 2008, p. S29 

Introduction: The constant scale (C) [41] is routinely used to assess the level 

<strong>of</strong> shoulder impairment during rehabilitation. C rates important issues, such as 

pain, range <strong>of</strong> <strong>motion</strong> and power. However, it does not comprise any item 

directly intended to follow the recovery <strong>of</strong> the normal coordination between 

humerus elevation and shoulder–girdle movements (named hereinafter girdle– 

humeral rhythm), which is an important aspect <strong>of</strong> rehabilitation. To fully 

understand the validity <strong>of</strong> C in monitoring the rehabilitation process, it is 

therefore essential to establish the relation betweenCscores and the 

compensatory movements affecting the girdle–humeral rhythm. In particular, 

the aim <strong>of</strong> this studywas to establish if: (1) a C score (CS) higher than 50/100 

and a humerus flexion score in C (HFS) higher than 8/10 can exclude the 

existence <strong>of</strong> compensatory movement; and (2) patients with overlapping CS 

and an identical HFS always present the same compensatory movements. 

Method: Five patients (P1–P5, 57±16 years-old) participated in this study. P1– 

P4 successfully underwent surgery for rotator cuff-tear and P5 for traumatic 

shoulder instability. All non-op sides were sound. Every week from the start <strong>of</strong> 

the active mobilization, each side <strong>of</strong> each patient was scored with C and its 3D 

kinematics was measured with the <strong>motion</strong> <strong>analysis</strong> protocol presented in 

paragraph 2.1.5 <strong>of</strong> this thesis: the assessment ended when the CS overcame 

50/100 and the HFS was higher than 8/10. In each <strong>motion</strong> <strong>analysis</strong> acquisition, 

subjects were asked to cyclically flex-extend the humerus in the sagittal plane 

for 5 times. The coordination plot relating the girdle elevation and the humerus 

flexion was then obtained for both sides. The affected and sound girdle– 

humeral rhythm were finally compared intra-subject by computing the mean 

difference (MD) between the girdle elevation <strong>of</strong> the sound side and that <strong>of</strong> the 

115

affected side, at 40◦, 80◦ and 100◦ <strong>of</strong> humerus flexion. 

Results and discussion: Table 1 reports the CSs and theMDvalues for each 

patient during the last assessment. From the data reported it can be concluded 

that a CS > 53/100 and a HFS > 8/10 do not exclude the existence <strong>of</strong> 

compensatory movement, since P1–P4 had a CS > 53/100 and a HFS > 8/10 

but showed marked compensatory movements, as proved by the MD values. 

Moreover, even though P4 and P5 had very close CSs and identical HFSs, they 

presented different patterns for the compensatory movements: no compensation 

for P5 while marked for P4. Different patterns were also found comparing P1 

and P2 (who showed HFS = 10/10), as well as for P5 and P3 (see www.inailstarter.org 

for images). In conclusion, patients with overlapping CSs and 

identical HFSs do not always present the same compensatory movements. The 

CS and HF scores considered were chosen since they are typical <strong>of</strong> patients 

after 3–5 weeks <strong>of</strong> active mobilization, i.e. in the second half <strong>of</strong> rehabilitation. 

From the results it can be concluded thatCshould be used with caution to draw 

conclusions on the recovery <strong>of</strong> shoulder normal kinematics. 

Patient ID P1 P2 P3 P4 P5 

CS 56 58 53 62 60 

HFS 10/10 10/10 8/10 8/10 8/10 

MD [deg] 

8.6; 8.1; 4.9; 5.4; 1; 9.4; -0.1 ;2.9 -5.4; - 

4.5 2.6 10.4 ;5.2 3.5; -0.7 

Table 1 - Constant scores (CS), HFS and MD for the 5 patients at the end <strong>of</strong> the assessment. 

116

2.1.2.2 

THE EVOLUTION OF COMPENSATION STRATEGIES IN 

TWO PATIENTS WITH SHOULDER INSTABILITY: A 

COMPARATIVE STUDY THROUGH QUANTITATIVE 

MOTION ANALYSIS 

Gar<strong>of</strong>alo P, Cutti AG, Filippi MV, Davalli A, Cappello A 

Gait & Posture, vol. 24, suppl. 1, November 2006, p. S39 

Abstract 

The aim <strong>of</strong> this work was to present the case study <strong>of</strong> two post-surgical patients 

treated for shoulder instability, assessed integrating the Constant impairment 

rating scale with a <strong>motion</strong> <strong>analysis</strong> protocol developed for monitoring the 

shoulder girdle compensatory strategies. In particular, for each patient the 

girdle-humerus coordination patterns <strong>of</strong> the impaired shoulder acquired before 

and after 2 weeks <strong>of</strong> active/active-assisted mobilisation 

were compared with that <strong>of</strong> the contralateral sound side. Results showed for 

one patient that the increased gleno-humeral RoM was followed by an 

increased shoulder girdle <strong>motion</strong> with abnormal patterns. This justified an 

adjustment <strong>of</strong> his rehabilitation program. 

The case study stressed therefore the need for a targeted <strong>analysis</strong> <strong>of</strong> 

compensation strategies 

for a deeper understanding <strong>of</strong> the therapeutic outcomes. 

Introduction 

As soon as patients who underwent surgery for a shoulder instability begin the 

active mobilisation, a reduced shoulder range <strong>of</strong> <strong>motion</strong> (RoM) and an altered 

scapulohumeral rhythm can be observed. In particular, an abnormal kinematics 

<strong>of</strong> the shoulder girdle (defined as the fictitious segment connecting the sternoclavicular 

with the gleno-humeral joint), is generally evident during arm 

elevation, involving range and/or timing <strong>of</strong> elevation-depression and 

protraction-retraction. One <strong>of</strong> the target <strong>of</strong> rehabilitation is therefore the 

117

ecovery <strong>of</strong> an adequate shoulder RoM while trying to remove these 

compensatory movements. To this regard, unfortunately, the usual clinical 

rating scales for assessment <strong>of</strong> shoulder impairment, i.e. Constant and ASES 

[41] , do not provide a full support to monitor the results <strong>of</strong> the rehabilitative 

intervention. In fact, even though recording information about the shoulder 

overall mobility, they do not address how a movement is performed, nor they 

describe the compensation strategy. 

Quantitative <strong>motion</strong> <strong>analysis</strong> with focused indexes extraction appears a possible 

solution to overcome these limitations [42,36]. The aims <strong>of</strong> this work were 

therefore: 1) to advance a simple <strong>motion</strong> <strong>analysis</strong> protocol for monitoring the 

girdle compensatory movements in patients with shoulder impairment; 2) to 

present the results obtained assessing two post-surgical patients treated for 

shoulder instability, integrating the Constant scale with the <strong>motion</strong> <strong>analysis</strong> 

protocol. In particular, the patients‘ different evolution <strong>of</strong> the shoulder girdlehumerus 

coordination with the therapy was discussed as well as its implications 

for the early adjustment <strong>of</strong> the physical therapy. 

Materials and Methods 

Motion <strong>analysis</strong> protocol 

The <strong>motion</strong> <strong>analysis</strong> protocol was intended to quantitatively assess, e.g. through 

a stereophotogrammetric system, the RoM and the coordination between 

shoulder girdle and 

humerus <strong>motion</strong> during 4 activities <strong>of</strong> the Constant scale: hand-behind-head 

and full shoulder flexion in the sagittal plane, handto-top-<strong>of</strong>-head and full 

shoulder adduction in the frontal plane. Each patient had to be acquired while 

performing the activities after 1 and 3 weeks <strong>of</strong> active/active-assisted 

mobilisation (AQ1, AQ2). Assuming the shoulder <strong>of</strong> one side <strong>of</strong> the body 

unimpaired (i.e. as the subject-specific gold-standard), in each acquisition each 

activity had to be repeated 5 times, firstly with the right and then with the left 

arm. From the biomechanical viewpoint, being interested in the orientation <strong>of</strong> 

the shoulder girdle with respect to the thorax and <strong>of</strong> the humerus with respect to 

the shoulder girdle, upper arms were modelled as open kinematic chains each 

formed by 3 segments (thorax in common, shoulder girdle and humerus), with 

5 degrees <strong>of</strong> freedom: 2 describing the mobility <strong>of</strong> the shoulder girdle [2], and 3 

<strong>of</strong> the glenohumeral joint (fig 1a). The thorax and humerus bone embedded 

systems <strong>of</strong> reference (SoR) were defined following the ISG recommendations 

118

[20]: H1 for the humerus, z axes pointing backward. For the (right) girdle, the x 

axis was assumed from IJ to GH, the z as the cross product <strong>of</strong> x and the y axis 

<strong>of</strong> the thorax and the y consequently. 

Through a static trial at the beginning <strong>of</strong> AQ1 and AQ2, the girdle and humerus 

SoRs were 

then reoriented in order to be parallel to that <strong>of</strong> the thorax when the subject 

stood in the upright position, arms along the body, neutral pronation: in this 

posture, therefore, all joint angles were zero. The axes <strong>of</strong> the left side SoRs 

were orientated to give rise to joint angles with the same sign <strong>of</strong> the right side. 

Joint angles were obtained decomposing the relative orientation <strong>of</strong> adjacent 

segments using appropriate sequences <strong>of</strong> Euler angles: protraction-retraction 

(PR-RE) and elevationdepression (EL-DE) for the shoulder girdle, flexionextension 

(FL-EX), ab-adduction (AB-AD) and internal-external rotation 

(INEX) or AB-AD, FL-EX and IN-EX for mostly sagittal or frontal tasks. In 

order to track the segments SoR during <strong>motion</strong> with a stereophotogrammetric 

system (Vicon 460, 6 cameras), 4 markers were placed on the thorax 

anatomical landmarks, while 4 on the humerus <strong>of</strong> each side to form a technical 

cluster (fig 1b). Anatomical calibrations were then performed as described in 

[43]. 

Figure 1 - Figure 1a,b (a) Open kinematic chain modelling the left arm; (b) marker-set used. 

Markers on IJ and PX [4] are not visible. Markers on acromia were used for visualisation purposes 

only. For both AQ1 and AQ2, for each activity and side, the RoM <strong>of</strong> shoulder girdle and 

glenohumeral joint were computed, as well as the angle-angle plots describing the coordination 

between the girdle degrees <strong>of</strong> freedom and the gleno-humeral ones [36,42]. Corresponding angleangle 

plots from AQ1 and AQ2 for the impaired and sound sides were then superimposed, to check 

for the evolution in the girdle-humerus coordination <strong>of</strong> the impaired shoulder in time and with 

respect to the unimpaired one. 

119

Patients description 

Two patients participated in the study, after signing an informed consent. A 30 

years old woman, initials S.Q., and a 21 years old man, initials E.G.. S.Q. 

underwent an arthroscopic surgery with debridement and plications at the left 

shoulder for a recent traumatic instability with condral lesions. E.G., instead, 

underwent an arthoscopic surgery at the left shoulder with capsula repairing 

and plications for an instability following recurrent episodes <strong>of</strong> scapulohumeral 

dislocation in the past 3 years. Both patients were acquired with the 

<strong>motion</strong> <strong>analysis</strong> protocol, as well as clinically assessed with the Constant (for 

impairment) and DASH [2] (for disability) rating scales the same day <strong>of</strong> the 

two acquisitions. The clinical questions we had to answer were if the active and 

active-assisted mobilisation followed by the patients was effectively 1) 

increasing the shoulder RoM, 2) restoring a girdle-humerus coordination 

consistent with the right unaffected side and 3) if the level <strong>of</strong> disability was 

decreasing. 

Results 

The Constant scales for the patients‘ impaired side in AQ1 and AQ2 are 

reported in table 3, 

evidencing a good improvement, mostly due to an increased RoM, and for S.Q. 

also due to a consistent decrease <strong>of</strong> pain. The DASH scale remained almost 

unchanged, decreasing from 25.83 to 24.16 for S.Q. and from 26.66 to 25.83 

for E.G., mostly due for both to the impossibility to practice the usual sport, to 

lift loads and to sleep without pain. The <strong>motion</strong> <strong>analysis</strong> acquisitions confirmed 

the overall RoM improvements <strong>of</strong> the gleno-humeral joint (table 1). In addition, 

the angle-angle plots brought in evidence the kinematic patterns <strong>of</strong> the girdle 

with respect to the humerus <strong>motion</strong>. Exemplificative <strong>of</strong> the subjects‘ behaviour 

were found the plots for the forward flexion task, which are reported in figure 

2. 

120

Table 1 - Constant scales for S.Q. and E.G. I1 and I2 refer to the impaired side in AQ1 and AQ2, 

respectively. U refers to the unimpaired side. 

Table 2 - Gleno-humeral RoM. For the hand-to-top-<strong>of</strong>-head task (HTH) the AB-AD and FL-EX 

angles are reported. A reduction in this latter indicates a more frontal (i.e. better) execution <strong>of</strong> the 

task. Sagittal hand-behind-head RoM is included in the forward elevation activity 

121

Figure 2 - Plots showing for the two subjects the coordination between the gleno-humeral flexion 

and girdle <strong>motion</strong> during the forward flexion activity. Shown are the overlapped patterns <strong>of</strong> the 

impaired side in AQ1 and AQ2 (denoted as I1 and I2 in the plots), compared to the unimpaired side 

for AQ2 (denoted as U), which was consistent with the one found for AQ1. For each acquisition 

and side, the patterns <strong>of</strong> the central 3 repetitions are reported, out <strong>of</strong> the 5 executed. 

Discussion and Conclusions 

Looking for a widely applicable and fast protocol, the acquisition <strong>of</strong> the single 

<strong>motion</strong> <strong>of</strong> clavicle and scapula was not considered, this requiring additional 

specialized instrumentation and generally being very time consuming [4,5]. 

Considering the definition <strong>of</strong> the SoRs, the reorientation <strong>of</strong> the girdle and 

humerus ones was performed to enhance the interpretability <strong>of</strong> the results, 

remaining more adherent to clinical expectations. The selection <strong>of</strong> a very 

limited number <strong>of</strong> tasks was forced by the limited time available for the 

acquisition, i.e. 30 minutes. However, in authors‘ opinion, those selected take 

into account the Constant scale activities where compensation strategies are 

more evident. Coming to the results, the DASH scale pointed out the 

unchanged disability level <strong>of</strong> the patients. This is reasonable given the absence 

in the therapy <strong>of</strong> force regaining exercises. Considering the exemplificative 

forward flexion task, an increased gleno-humeral RoM could be observed for 

both subjects consistently with Constant. However, E.G. showed, between the 

two acquisitions, very similar coordination patterns for the impaired side, 

122

which remained far from that <strong>of</strong> the sound side. Moreover, the increased RoM 

in AQ2 impaired pattern was followed by an increased girdle elevation, and by 

a PR-RE pattern which drifted away from the sound. 

Similar results were obtained from the other activities, even if not observing a 

worsening <strong>of</strong> the PR-RE. On the contrary, S.Q. patterns in AQ2 went closer to 

that <strong>of</strong> the healthy side, with an increased gleno-humeral RoM and a decreased 

EL-DE and PR-RE. A moderate delay in girdle activation appeared in AQ2, 

which was confirmed by the other activities. From the results obtained it can be 

stated therefore that, although not still decreasing the disability level, the 

therapy applied on S.Q. has already obtaining positive effects both on RoM and 

compensatory movements, while on E.G. it increased the RoM while increasing 

the entity <strong>of</strong> the girdle abnormal <strong>motion</strong>. Since E.G. came from a 3 years 

history <strong>of</strong> shoulder dislocation, longer terms results are expected. However, an 

increment in the execution <strong>of</strong> proprioceptive and scapulo-humeral coordination 

exercises could be recommended. It can be concluded therefore that the <strong>motion</strong> 

<strong>analysis</strong> protocol have positively integrated the Constant scale, explicating the 

relation between a general increment in gleno-humeral RoM, observed in both 

patients, and the underlying girdlehumerus coordination, with implications for 

the rehabilitation program <strong>of</strong> one <strong>of</strong> them. 

123

2.1.2.3 

THE RELATION BETWEEN SCAPULAR AND SHOULDER-GIRDLE 

MOTION IN SUBJECTS WITH AND WITHOUT SHOULDER 

IMPINGEMENT 

Gar<strong>of</strong>alo P. 1,2 , Cutti A.G. 1 , Ludewig P. 3 , Phadke V. 3 , Cappello A. 2 

1 INAIL Prostheses Centre, Vigorso di Budrio, Italy 

2 DEIS, University <strong>of</strong> Bologna, Italy 

3 Program in Rehabilitation Science, University <strong>of</strong> Minnesota, Minneapolis, 

MN, USA 

Proc. ISG 2008, 10-13 July, 2008, Bologna, Italy 

Abstract 

A protocol was developed to measure the shoulder-girdle elevation-depression 

(GED) and protraction-retraction in clinical settings, when scapula tracking is 

not feasible. The importance <strong>of</strong> the scapula however remains.The questions we 

tried to answer in this preliminary study were therefore: 1) which is the relation 

between GED and scapula medio-lateral rotations (SML) in asymptomatic 

subjects, during shoulder forward flexion and ab-adduction 2) does this 

relation change in subjects with shoulder impingement 3) looking at the GED 

and SML, is it possible to distinguish between subjects with and without 

impingement Results show that for asymptomatic subjects there is high 

correlation (r>0.93) between GED and SML, which decreases to low or 

moderate for symptomatic subjects. Moreover, a clustering <strong>of</strong> subjects appears 

possible considering as predictors: 1) the correlation <strong>of</strong> GED and SML during 

shoulder forward flexion, and 2) the ratio between the RoM <strong>of</strong> SML and the 

RoM <strong>of</strong> GED, during shoulder ab-adduction. 

1. Introduction 

Alterations <strong>of</strong> the scapulohumeral rhythm <strong>of</strong>fer important indications <strong>of</strong> 

shoulder pathologies. Unfortunately, the measurement <strong>of</strong> scapulothoracic 

rotations is not always possible, e.g. due to clinical routine constraints which 

preclude the use <strong>of</strong> currently available tracking systems for the scapula [1-2]. 

The need to complete the acquisitions <strong>of</strong>: 1) both shoulders <strong>of</strong> a subject, 2) 

124

either muscular or lean, 3) within 30 minutes, 4) with no constraint on the 

maximal humeral elevation, is a concrete example <strong>of</strong> such a combination <strong>of</strong> 

constraints, taken from the routine <strong>of</strong> the INAIL Prostheses Centre. To face this 

situation, a <strong>motion</strong> <strong>analysis</strong> protocol was developed, which allows us to easily 

measure the overall <strong>motion</strong> <strong>of</strong> the shoulder-girdle (defined as the fictitious 

segment connecting the thorax with the glenohumeral joint), and to relate its 

<strong>motion</strong> to humeral rotations. The protocol, therefore, allows us to measure the 

girdlehumeral rhythm. The protocol was tested [3-5], and gave pro<strong>of</strong> <strong>of</strong> 

effectiveness in monitoring the evolution <strong>of</strong> shoulder-girdle compensatory 

movements in post-op patients treated for rotator-cuff tears and glenohumeral 

instability. 

The importance <strong>of</strong> scapula rotations, however, still remains. The following 

question therefore emerges: given the fact that girdle rotations can be easily 

tracked, while scapular rotations can not, what can be said about the 

scapulohumeral rhythm looking at the girdlehumeral rhythm More 

specifically, the aim <strong>of</strong> this preliminary investigation was to answer to the 

following questions: 

1) during humeral forward flexions (FLEX) and abductions (ABAD), which is 

the relation between the scapula medio-lateral rotation (SML) and the girdle 

elevation-depression (GED), in asymptomatic subjects (AS); 

2) does this relation change in Symptomatic Subjects (SS) with shoulder 

impingement 

3) can the relation be used to classify subjects as SS or AS 

2. Material and methods 

2.1 Subject 

Seven AS and 3 SS participated in this study, after giving their informed 

consent. Subjects were recruited and evaluated by the authors <strong>of</strong> the University 

<strong>of</strong> Minnesota, in Minneapolis. Physical examinations confirmed or excluded 

shoulder impingement. However, the subject classified as #2 subsequently 

developed mild impingement symptoms at a later date; for this reason subject 

#2 was reclassified for this work as symptomatic. The 6 AS subjects were on 

average 29±6 years-old, while the 4 SS were on average 33±12 years-old. 

2.2 Motion Analysis Protocol 

Looking for the comparison between GED and SML the kinematics <strong>of</strong> the 

scapula and girdle with respect to the thorax were measured. The girdle- 

125

thoracic <strong>motion</strong> was modelled as a two-degree <strong>of</strong> freedom (DoF) rotational 

joint [5]: elevation-depression (ED) and protraction-retraction. Scapulothoracic 

<strong>motion</strong> was modelled as a 3 DoF joint: protraction-retraction, mediolateral 

rotation (ML) and axial rotation. The ISG standard [6] was followed for 

the definition <strong>of</strong> the anatomical frames <strong>of</strong> thorax and scapula, while [5] was 

followed for the definition <strong>of</strong> the girdle anatomical frame. A Flock-<strong>of</strong>-birds 

system (Ascension, US) was used for the acquisitions. Sensors were attached to 

each bony segment via a 2.5mm diameter pin placed into the bones bicortically. 

A surface sensor was placed over the sternum. Anatomical landmarks were 

calibrated with respect to the sensors and anatomical frames were found 

consequently for each frame <strong>of</strong> the acquisitions. Subjects were asked to execute 

twice a FLEX and a ABAD. 

2.2 Data <strong>analysis</strong> 

For each patient, both for FLEX and ABAD, GED and SML were analyzed. In 

particular, we computed, 1) the ratio between the RoM <strong>of</strong> SML and GED 

(S/G), and 2) the correlation coefficient (r) between GED and SML. After 

checking for normal distributions, the mean and std <strong>of</strong> r and S/G was computed 

for the AS group and the SS group. Differences between the means for AS and 

SS were analyzed either with a 1-way ANOVA or through the Kruskal-Wallis 

test. 

3. Results 

Figs. 1-2 report scatter plots for the subjects, with r on the x-axis and S/G on 

the y-axis, for FLEX and ABAD, respectively. For FLEX (fig. 1), r between 

GED and SML for AS was high, while it was low for SS. In particular, r mean 

value for AS was 0.93±0.05, while it was 0.60±0.30 for SS. The p-value 

between the two groups was p=0.059, which does not indicate a statistically 

significant difference, but indicates a good trend <strong>of</strong> separation between AS and 

SS. For FLEX, the ratio S/G was lower for AS (3.69±0.81) compared to SS 

(4.65±2.46). The p-value between the two groups was however rather low 

(p=0.38). For ABAD (fig. 2), r between GED and SML for AS was high 

(0.95±0.05), and it was moderate for SS (0,79±0,23). The p-value between the 

groups was p=0.13. The ratio S/G for AS was 3.62±0.76 while for SS it was 

4.57 ±0.92 (p between groups = 0.11). This can provide some indications <strong>of</strong> a 

stiffer girdle in SS compared to AS. For the classification <strong>of</strong> subjects <strong>based</strong> on r 

and S/G and considering fig. 1, r for FLEX divides SS from AS, with the only 

126

exception <strong>of</strong> subject #10 (SS) (fig. 1). #10 was however markedly distant from 

all other subjects observing the ratio S/G in fig. 2 (ABAD). Combining r for 

FLEX and S/G for ABAD, it is possible to isolate the AS from the SS, resulting 

in a clustering <strong>of</strong> the subjects (fig. 3). 

4. Discussion and conclusions 

The aim <strong>of</strong> this study was to answer to 3 questions concerning the relation 

between GED and SML. Considering the first question (relation between GED 

and SML in AS) we can conclude that for AS a high correlation exists between 

the two angles (r>0.93, on average) for both FLEX and ABAD. This is 

consistent with patterns previously reported for the clavicle and scapula [7-8]. 

Moreover, for AS the RoM <strong>of</strong> GED is on average 3.6 times less than the RoM 

<strong>of</strong> SML for both movements. Considering the second question (modification <strong>of</strong> 

the relation between GED and SML in SS) results show the tendency <strong>of</strong> r to 

decrease from AS to SS during FLEX. Moreover, AS seem to elevate more the 

girdle compared to SS, but this preliminary indication will have to be 

confirmed on a broader population, and supported by statistically significant 

results. Finally, considering the third question (clustering <strong>of</strong> SS and AS) results 

show that combining r during FLEX with S/G in ABAD it may be possible to 

obtain a clustering <strong>of</strong> subjects. It was particularly interesting to notice that the 

clustering indicated subject #2 as SS, even though symptoms emerged later in 

time. Further <strong>analysis</strong> are required to draw definitive conclusions, but the 

possibility to cluster subjects <strong>based</strong> on simple variables appears interesting 

from a clinical perspective. 

127

5. References 

[1] Johnson, G.R. et al. (1993) Clin Biomech, 8, 269-273 

[2] Ludewig, P. et al. (2000) Phys Ther, 80:276-291 

[3] Cutti, A.G. et al. (2006) Proc ISG 2006 

[4] Gar<strong>of</strong>alo, P. et al. (2007) J. Biomech, 40, S106 

[5] Gar<strong>of</strong>alo, P. et al. (2008) Proc. ISG 2008 

[5] Wu, G. et al. (2005) J. Biomech, 38, 981-992 

[6] McClure P.W. et al, (2001) JSES, 10:269-77 

[7] van der Helm, F. et al. (1995) J Biomech Eng, 117, 27-40 

128

2.1.3 References 

1. Rockwood C, Matsen IF, Wirth M. The Shoulder. Saunders, USA; 2004. 

2. Finch E, Brooks D, Statford P. Physical Rehabilitation Outcome Measures. 

Lippincott Williams & Wilkins, Canada; 2002. 

3. Jurriaan H de Groot. The variability <strong>of</strong> shoulder <strong>motion</strong>s recorded by means 

<strong>of</strong> palpation. Clin Biomech (Bristol, Avon). 1997;12(7):461-472. 

4. Johnson G, Stuart P, Mitchell S. A method for the measurement <strong>of</strong> threedimensional 

scapular movement. Clinical Biomechanics. 1993;8(5):269-273. 

5. Karduna AR, McClure PW, Michener LA, Sennett B. Dynamic 

measurements <strong>of</strong> three-dimensional scapular kinematics: a validation study. J 

Biomech Eng. 2001;123(2):184-190. 

6. Ludewig PM, Cook TM. Alterations in shoulder kinematics and associated 

muscle activity in people with symptoms <strong>of</strong> shoulder impingement. Phys Ther. 

2000;80(3):276-291. 

7. D.J. Magermans, E.K.J. Chadwick, H.E.J. Veeger, F.C.T. van der Helm. 

Requirements for upper extremity <strong>motion</strong>s during activities <strong>of</strong> daily living. Clin 

Biomech (Bristol, Avon). 2005;20(6):591-599. 

8. Marchese S, Johnson G. Measuring the kinematics <strong>of</strong> the clavicle. 6th 

International Symposium on the 3D Analysis <strong>of</strong> Human Movement. Cape Town, 

South Africa. 2000:37-40. 

9. BS AGM, MS SL, BS PG, et al. Effect <strong>of</strong> rotator cuff pathology on shoulder 

rhythm. Journal <strong>of</strong> Shoulder and Elbow Surgery. 2005;14(1):S58-S64. 

10. Meskers C, Vermeulen H, de Groot J, van Der Helm F, Rozing P. 3D 

shoulder position measurements using a six-degree-<strong>of</strong>-freedom electromagnetic 

tracking device. Clin Biomech (Bristol, Avon). 1998;13(4-5):280-292. 

11. Pascoal AG, van der Helm FF, Pezarat Correia P, Carita I. Effects <strong>of</strong> 

129

different arm external loads on the scapulo-humeral rhythm. Clin Biomech 

(Bristol, Avon). 2000;15 Suppl 1:S21-24. 

12. Roy J, M<strong>of</strong>fet H, Hébert LJ, St-Vincent G, McFadyen BJ. The reliability <strong>of</strong> 

three-dimensional scapular attitudes in healthy people and people with shoulder 

impingement syndrome. BMC Musculoskelet Disord. 2007;8:49. 

13. van Andel CJ, Wolterbeek N, Doorenbosch CAM, Veeger DHEJ, Harlaar J. 

Complete 3D kinematics <strong>of</strong> upper extremity functional tasks. Gait Posture. 

2008;27(1):120-127. 

14. van der Helm FC, Pronk GM. Three-dimensional recording and description 

<strong>of</strong> <strong>motion</strong>s <strong>of</strong> the shoulder mechanism. J Biomech Eng. 1995. Available at: 

http://www.biomedsearch.com/nih/Three-dimensional-recording-description<strong>motion</strong>s/7609482.html. 

15. Carel G.M. Meskers, Michiel A.J. van de Sande, Jurriaan H. de Groot. 

Comparison between tripod and skin-fixed recording <strong>of</strong> scapular <strong>motion</strong>. J 

Biomech. 2007;40(4):941-946. 

16. Cutti A, Gar<strong>of</strong>alo P, Filippi M. The evolution <strong>of</strong> compensation strategies in 

two patients with shoulder instability: a comparative study through quantitative 

<strong>motion</strong> <strong>analysis</strong>. Proc ISG2006 Chicago, USA. 2006. 

17. Stagni R, Fantozzi S, Cutti A, Cappello A. KINEMATIC ANALYSIS 

TECHNIQUES AND THEIR APPLICATION IN BIOMECHANICS. In: 

BIOMECHANICAL SYSTEMS TECHNOLOGY — A 4-Volume Set.Vol 3: 

Muscular Skeletal Systems.; :135-183 . 

18. P. Gar<strong>of</strong>alo, A.G. Cutti, M.V. Filippi, et al. Limitations <strong>of</strong> the constant scale 

in the assessment <strong>of</strong> shoulder compensatory strategies. Gait Posture. 

2008;28:S29. 

19. Lorenzo Chiari, Ugo Della Croce, Alberto Leardini, Aurelio Cappozzo. 

Human movement <strong>analysis</strong> using stereophotogrammetry: Part 2: Instrumental 

errors. Gait Posture. 2005;21(2):197-211. 


130



2005;38(5):981-992. 

21. Veeger HE, Helm FVD, Rozendal R.H. Orientation <strong>of</strong> the scapula in a 

simulated wheelchair push. Clinical Biomechanics. 1993;8(2):59-111. 

22. Cappozzo A, Catani F, Leardini A, Benedetti MG, Croce UD. Position and 

orientation in space <strong>of</strong> bones during movement: experimental artefacts. Clin 

Biomech (Bristol, Avon). 1996;11(2):90-100. 

23. Cutti A, Paolini G, Troncossi M. A new technique for compensating the 

s<strong>of</strong>t tissue artefact at the upper-arm: in vitro validation. J Mech Med Biol . 

2005;5:1-15. 

24. Murray IA, Johnson GR. Definition <strong>of</strong> marker positions and technical 

frames for studying the kinematics <strong>of</strong> the shoulder. Proceedings <strong>of</strong> Third 

Conference <strong>of</strong> the International Shoulder Group, Newcastle. 2000. 

25. Charlton I. A Model for the Prediction <strong>of</strong> the Forces at the Glenohumeral 

Joint. Ph.D. Thesis, University <strong>of</strong> Newcastle upon Tyne. 2003. 

26. Justin J. Kavanagh, Steven Morrison, Daniel A. James, Rod Barrett. 

Reliability <strong>of</strong> segmental accelerations measured using a new wireless gait 

<strong>analysis</strong> system. J Biomech. 2006;39(15):2863-2872. 

27. Kadaba MP, Ramakrishnan HK, Wootten ME, et al. Repeatability <strong>of</strong> 

kinematic, kinetic, and electromyographic data in normal adult gait. J. Orthop. 

Res. 1989;7(6):849-860. 

28. Lee RYW, Laprade J, Fung EHK. A real-time gyroscopic system for threedimensional 

measurement <strong>of</strong> lumbar spine <strong>motion</strong>. Medical Engineering & 

Physics. 2003;25(10):817-824. 

29. Ruth E. Mayagoitia, Anand V. Nene, Peter H. Veltink. Accelerometer and 

rate gyroscope measurement <strong>of</strong> kinematics: an inexpensive alternative to 

optical <strong>motion</strong> <strong>analysis</strong> systems. J Biomech. 2002;35(4):537-542. 

131

30. Peter M. Mills, Steven Morrison, David G. Lloyd, Rod S. Barrett. 

Repeatability <strong>of</strong> 3D gait kinematics obtained from an electromagnetic tracking 

system during treadmill loco<strong>motion</strong>. J Biomech. 2007;40(7):1504-1511. 

31. Yavuzer G, Oken O, Elhan A, Stam HJ. Repeatability <strong>of</strong> lower limb threedimensional 

kinematics in patients with stroke. Gait Posture. 2008;27(1):31-35. 

32. Weir JP. Quantifying test-retest reliability using the intraclass correlation 

coefficient and the SEM. J Strength Cond Res. 2005;19(1):231-240. 

33. Stergiou N. Innovative analyses <strong>of</strong> human movement. Human Kinetics 

Publishers; 2003. 

34. Bland M. An introduction to medical statistics. Oxford University Press, 

USA; 2000. 

35. Hopkins WG. Measures <strong>of</strong> reliability in sports medicine and science. Sports 

Med. 2000;30(1):1-15. 

36. McClure PW, Michener LA, Sennett BJ, Karduna AR. Direct 3- 

dimensional measurement <strong>of</strong> scapular kinematics during dynamic movements 

in vivo. J Shoulder Elbow Surg. 2001;10(3):269-277. 

37. Borstad JD, Ludewig PM. Comparison <strong>of</strong> scapular kinematics between 

elevation and lowering <strong>of</strong> the arm in the scapular plane. Clin Biomech (Bristol, 

Avon). 2002;17(9-10):650-659. 

38. Lenh<strong>of</strong>f MW, Santner TJ, Otis JC, et al. Bootstrap prediction and 

confidence bands: a superior statistical method for <strong>analysis</strong> <strong>of</strong> gait data. Gait & 

Posture. 1999;9(1):10-17. 

39. Michael H. Schwartz, Joyce P. Trost, Roy A. Wervey. Measurement and 

management <strong>of</strong> errors in quantitative gait data. Gait Posture. 2004;20(2):196- 

203. 

40. Zoubir A, Boashash B. The bootstrap and its application in signal 

processing. Signal Processing Magazine, IEEE. 1998;15(1):56-76. 

132

41. Pynsent P. Outcome measure in orthopaedics. Butterworth–Heinemann 

Ltd. 

42. Klopčar N, Lenarčič J. Bilateral and unilateral shoulder girdle kinematics 

during humeral elevation. Clinical Biomechanics. 21:S20-S26. 

43. Cutti A, Paolini G, Troncossi M, Cappello A. Skin artifact evaluation in 

humeral axial rotation. Gait & Posture. 2003;18:5. 

133


2.2.1 Development <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol for the kinematics 

<strong>of</strong> upper-limb myoelectric prostheses 

BIOMECHANICAL ANALYSIS OF AN UPPER LIMB 

AMPUTEE AND HIS MYOELECTRIC PROSTHESIS: A 

CASE STUDY 

Cutti AG, Gar<strong>of</strong>alo P, Janssens K, Davalli A, Sacchetti R 

Orthopaedie Technik (Quarterly), 2007, Issue 1, 6-15 

Abstract 

The aim <strong>of</strong> this study was to address through quantitative techniques two 

interconnected problems in myoelectric prosthesis <strong>development</strong> and use: on one 

hand, the early testing <strong>of</strong> innovative prostheses in real-life conditions providing 

manufacturers with quantitative information regarding the prostheses 

performance and reliability; on the other, the increasing need for quantitative 

assessment <strong>of</strong> patients ability in myoelectric control, to facilitate the 

cooperative work <strong>of</strong> CPOs in tuning the prosthesis and <strong>of</strong> OTs in focus training. 

The method proposed is <strong>based</strong> on quantitative <strong>motion</strong> <strong>analysis</strong>, biomechanical 

modelling and acquisition <strong>of</strong> the prosthesis EMG control signals. It was applied 

in-vivo for testing the performances and reliability <strong>of</strong> the pre-market version <strong>of</strong> 

the new Otto Bock Dynamic Arm elbow, as well as the ability in myoelectric 

control <strong>of</strong> an experienced transhumeral amputee to whom the prosthesis was 

fitted. Results showed that the elbow maximum flexion and mean velocity, 

ranging between 139°- 164°/s (for no load lifted by the elbow) to 118°-24°/s 

(while lifting 3 kg), appear adequate for the great majority <strong>of</strong> activities <strong>of</strong> the 

daily living, while preserving patient‘s safety. Occasional un responsiveness <strong>of</strong> 

the elbow when the hand was carrying loads was reported to the manufacturer: 

for the prosthesis currently on the market OttoBock changed the calibration 

procedure <strong>of</strong> the elbow load cells to fix the problem. Specific exercises for 

myoelectric control assessment showed the ability <strong>of</strong> the patient to take 

advantage <strong>of</strong> the proportional control while highlighting difficulties in 

134

generating valid cocontractions. In authors‘ opinion the integration <strong>of</strong> the 

proposed method with clinical rating scales appears very promising. Current 

effort are focused on the substitution <strong>of</strong> the expensive stereophotogrammetric 

system for <strong>motion</strong> <strong>analysis</strong> with cheep sensors <strong>based</strong> on accelerometers and 

gyroscopes and on a simplification <strong>of</strong> the interconnections required for 

recording the EMG control signals. 

135


When treating upper extremity amputees, occupational therapy should enable 

patients to actively use the prosthesis, allowing for unrestricted possibilities in 

the execution <strong>of</strong> patient-relevant activities <strong>of</strong> the daily living (ADLs) and 

participation in society [13]. If the amputee is fitted with a myoelectric 

prosthesis, a fundamental pre-requirement for a full exploitation <strong>of</strong> the device is 

the person‘s capacity for myoelectric control [7]. For above-elbow amputees 

fitted with motorized elbow, wrist and hand, this translates into an appropriate 

modulation <strong>of</strong> muscle contraction for speed and force control during elbow 

flexion-extension, wrist prono-supination and hand opening-closing. Moreover, 

good control requires fast and precise switching between the different motors. 

As a guide for the therapy, tuning <strong>of</strong> the prosthesis and for monitoring patients 

advancements, the adoption <strong>of</strong> tools for the assessment <strong>of</strong> control capacity 

appears essential [15]. In this context quantitative <strong>motion</strong> <strong>analysis</strong>, 

biomechanical modeling and acquisition <strong>of</strong> the EMG signals used for prosthesis 

control can provide a quantitative and objective description <strong>of</strong> the inter action 

between the amputee and his/her prosthesis during selected ADLs or specific 

control exercises. On one hand, in fact, through quantitative <strong>motion</strong> <strong>analysis</strong> 

and biomechanical model ling <strong>of</strong> the anatomical/artificial limbs [3] it is 

possible to acquire the 3D joint kinematics <strong>of</strong> the patient and prosthesis. On the 

other, the acquisition <strong>of</strong> the EMGs control signals allows the OT/CPO to 

quantitatively observe the patient EMG modulation ability and the errors he/she 

makes in switching between motors. Moreover, if the subject acquired with 

these quantitative techniques is an amputee fitted with a new prosthetic arm, the 

information provided can be useful not only for the OT/CPO but also for the 

prosthesis designer: the in-vivo results obtained for the prosthesis from 

quantitative <strong>motion</strong> <strong>analysis</strong> can in fact be easily compared with those from invitro 

tests, bringing in evidence potential difference due to real-life control 

signals, thus allowing for early adjustments. 

The aim <strong>of</strong> this work is to give an example <strong>of</strong> this clinical/technological 

assessment, presenting the results obtained for a transhumeral amputee fitted 

with the pre-market version <strong>of</strong> the new OttoBock 12K100 Dynamic Arm, a 

myoelectrically controlled and electromotive powered elbow. Firstly, the range 

<strong>of</strong> <strong>motion</strong>, velocity and reliability <strong>of</strong> the Dynamic Arm when controlled in-vivo 

by the patient‘s EMG signals were quantitatively assessed. In addition, the 

movements at the shoulder required to the patient to execute an elbow flexion- 

136

extension task were measured in order to monitor possible compensatory 

movements. Finally, the ability <strong>of</strong> the patient in controlling the prosthesis 

during specific assessment exercises was quantitatively evaluated. These 

quantitative data were completed by qualitative information obtained from the 

patient by means <strong>of</strong> a questionnaire, collecting his feelings about pros and cons 

<strong>of</strong> the prosthesis and its impact on quality <strong>of</strong> living. 

Material and Methods 

The Otto Bock Dynamic Arm 

Described by Otto Bock as a milestone in the <strong>development</strong> <strong>of</strong> microprocessorcontrolled 

prosthetics systems for the elbow [11], the 12K100 Dynamic Arm is 

a myoelectrically controlled and electromotive powered elbow joint for the 

fitting <strong>of</strong> upper arm amputees up to very distal trans-humeral amputation levels. 

The main advantages <strong>of</strong> the Dynamic Arm are: 1) high lifting and holding force 

(with lift arm set to 235 mm: lifting force=60 N; holding force=230 N); 2) 

minimal lift time 0.5 s for a forearm length <strong>of</strong> 235 mm and hand size 7; 3) 

natural free swing characteristics; 4) extended control possibilities for the 

amputee: repositioning <strong>of</strong> the elbow without an ―unlock signal‖, proportional 

control <strong>of</strong> the (optional) wrist rotator in both directions, simultaneous control <strong>of</strong> 

elbow and hand, proportional speed and grip control <strong>of</strong> the terminal device; 5) 

the Dynamic Arm and hand/ terminal device prosthesis can be controlled by 

electrodes, linear control elements, switches or a combination. 

137

Figure 1a,b - Frontal (a) and sagittal (b) view <strong>of</strong> the residual limb. 

Subject description 

The subject was a 52 years-old Caucasian Italian male, initials M. G., who 

sustained a traumatic amputation (type: myoplastik) <strong>of</strong> the left upper arm at 

high level (residual length: 35%) due to a work-related injury in 1970 

(Fig.1a,b). Upon an unsuccessful reimplantation, he was fitted with his first 

prosthesis in October 1971. M. G. actively uses the prosthesis all day long. His 

medical history includes no other severe trauma or disorders. The patient 

signed a written informed consent prior to all testing. 

Prosthesis set-up 

Prior to the Dynamic Arm, the patient was equipped with a hybrid prosthesis, 

i.e. myoelectric control <strong>of</strong> the hand and wrist by means <strong>of</strong> biceps and triceps 

signals with kinematic control <strong>of</strong> the elbow. On October 2005, the subject was 

fitted with the Dynamic Arm, ensuring him with myoelectric control <strong>of</strong> hand 

and wrist as well as the elbow joint (Fig. 2a,b). Lever arm length, 

corresponding to the distance from the rotational axis <strong>of</strong> the elbow to the centre 

138

<strong>of</strong> the grasping area, was 368 mm. Electrodes were placed over the residual 

muscle bellies <strong>of</strong> the biceps and triceps. The prosthesis control strategy was setup 

through the specialized ElbowS<strong>of</strong>t s<strong>of</strong>tware provided by Otto Bock with the 

Dynamic Arm. A short cocontraction (CoCo) <strong>of</strong> both muscle groups enabled a 

switching between motors in the cyclic sequence hand-elbow-wrist. 

Figure 2 - Frontal (a) and sagittal (b) view <strong>of</strong> patient wearing the Dynamic Arm 

Con traction <strong>of</strong> the biceps resulted in flexion <strong>of</strong> the elbow, closing <strong>of</strong> the hand 

or pronation <strong>of</strong> the wrist, depending on the selected function, whilst activation 

<strong>of</strong> the triceps resulted in the opposite movements. A muscle contraction was 

adequate to activate a selected <strong>motion</strong> only if it overcame the prosthesis ON 

threshold, i.e. 0.54 V; the <strong>motion</strong> was deactivated when the signals went below 

the OFF threshold, i.e. 0.38 V. A CoCo was recognized only if both signals, 

after overshooting the correspondent CoCo threshold within 80 ms from the 

first crossing <strong>of</strong> the ON threshold, went below the OFF within the time set with 

the ElbowS<strong>of</strong>t CoCo timer, here 800 ms. The CoCo threshold was set equal to 

the ON threshold. The velocity <strong>of</strong> the elbow was set to its 84%, for safety 

reasons. A tactile signal, i.e. one, two or three short vibrations, informed the 

subject on a successful switch to elbow, wrist or hand function. 

139

Figure 3 - Block diagram <strong>of</strong> the EMG acquisition system. 

Quantitative assessment <strong>of</strong> devices and biomechanical model 

For the assessment <strong>of</strong> the prosthesis and patient, quantitative information 

regarding the control EMG signals and kinematics were obtained. For the 

acquisition and data storage <strong>of</strong> the former, we developed a hardware and 

s<strong>of</strong>tware in Visual Basic. The EMGs recorded were those received by the 

prosthesis, since these were extracted in parallel from the wire connecting the 

electrodes to the Dynamic Arm. The hardware converted the analogic signals 

into digital and sent the data to the s<strong>of</strong>tware by wire (RS 232 protocol) or via 

Bluetooth connection (Fig. 3). For prosthesis and human <strong>motion</strong> acquisition a 

stereophotogrammetric system was used (Vicon 460, Oxford Metrics, UK), 

synchronized via-hardware with the EMG recorder. Being interested in elbow, 

wrist and hand prosthesis <strong>motion</strong>s as well as in the patient‘s shoulder and neck 

movements, 20 retro reflective markers were attached on head, thorax, socket, 

forearm, hand (Fig. 4a). These segments form an open kinematic chain with 11 

degrees <strong>of</strong> freedom (10 active and 1 passive), associated to the neck, shoulder 

girdle, gleno-humeral joint, elbow, wrist and hand (Fig. 4b). To define the 

mobility <strong>of</strong> the first 5 joints, at least one system <strong>of</strong> reference (SoR) had to be 

defined for each segment forming the joints. For the head, thorax, socket and 

proximal humerus rus (defined as the part <strong>of</strong> the humerus which forms the 

glenohumeral joint) these were obtained through the ―calibration‖ <strong>of</strong> relevant 

anatomical/prosthetic landmarks with respect to the correspondent cluster <strong>of</strong> 

markers [4,6,9,1]. For the definition <strong>of</strong> the distal humerus and forearm SoR, we 

combined the use <strong>of</strong> well-identifiable landmarks, with a functional, 

140

optimization-<strong>based</strong> method [14], which enables to compute the real axis <strong>of</strong> 

rotation <strong>of</strong> the elbow (flexion-extension) and <strong>of</strong> the wrist (prono-supination) 

[12]. Joint angles were then obtained decomposing the relative orientation <strong>of</strong> 

adjacent segments using appropriate sequences <strong>of</strong> Euler angles: flexionextension, 

lateral flexion (LF) and axial rotation for the neck, protractionretraction 

(PR-RE) and elevation-depression (EL-DE) for the shoulder girdle, 

flexion-extension (FL-EXsh), abduction-adduction (AB-ADsh) for the shoulder 

(the rotation order depends on the specific task), flexion-extension (FL-EXel) 

and pronation-supination (PS) for the elbow. In addition, in order to describe 

the hand opening and closing movement during pros thesis control tasks, we 

completed the kinematic chain computing the distance dHand between two 

markers placed on the thumb and the index finger (Fig. 4b). 

Fig. 4a, b Marker-set used for <strong>motion</strong> <strong>analysis</strong> (a): markers on right and left acromion are used for 

visualization purposes only; a marker is also placed on the 8th thoracic vertebra (not visible in the 

figure); kinematic model <strong>of</strong> the amputee with the prosthesis (b). 

141

Motion <strong>analysis</strong> protocol 

Elbow kinematic and reliability assessment 

In order to evaluate the prosthetic elbow kinematics and reliability, M. G. was 

acquired with the stereophotogrammetric and EMG systems during the 

execution <strong>of</strong> 6 elbow flexion-extension trials, with different loads on the hand: 

0 kg, 0.5 kg, 1 kg, 1.5 kg, 2 kg, 3 kg. For every trial, at least three consecutive 

flexion-extension cycles were repeated. In every cycle, the flexion (extension) 

was isolated: a flexion (extension) started from the activation <strong>of</strong> the biceps 

(triceps) EMG signal (over shooting <strong>of</strong> the ON threshold) and ended at the zero 

crossing <strong>of</strong> the FL-EXel velocity. The latency <strong>of</strong> the prosthesis from EMG 

activation to the start <strong>of</strong> <strong>motion</strong> was thus embedded into the assessment and 

also separately measured. Among the flexions and extensions we distinguished 

two different behaviours <strong>of</strong> the elbow, hereinafter referred to as ―optimal‖ and 

―failed‖. An ―optimal‖ flexion (extension) was defined as that in which the 

elbow movement followed a valid EMG signal activation. A ―failed‖ flexion 

(extension) was defined as that in which the elbow movement could not be 

obtained although a correct activation <strong>of</strong> the biceps (triceps) EMG signal was 

generated. Since OttoBock informed the authors that EMG activations over 

threshold but shorter than the latency for each load are automatically neglected 

by the prosthesis control system, only flexions/extensions where the EMG 

signals were over threshold and longer than the maximum latency measured in 

optimal flexions/extensions were considered. As a first index to assess the 

Dynamic Arm reliability, we compared the number <strong>of</strong> optimal and failed 

flexions and extensions. In addition, considering only the failed flexions 

(extensions), we further characterised the reliability by pointing out the number 

<strong>of</strong> failed EMG activations required before obtaining the <strong>motion</strong>. The optimal 

performances were instead analysed by computing the elbow range <strong>of</strong> <strong>motion</strong>, 

maximum flexion, time to reach the maximum flexion, mean and peak rotation 

velocities for each load. 

Shoulder assessment during elbow flexion-extension 

Based on the results obtained with a previous case study [5], we analysed the 

trials reported above by studying the range <strong>of</strong> <strong>motion</strong> and angle patterns at the 

glenohumeral joint, shoulder girdle and head lateral flexion <strong>of</strong> the patient while 

controlling the prosthesis. This was done in order to monitor possible 

142

compensatory movements. 

Exercises for assessing ability and strategy in prosthesis control 

To help assessing the ability <strong>of</strong> the patient in controlling the prosthesis we 

conceived the following three exercises: 

1. fine hand control: the subject was instructed to pinch grasping a cylindrical 

object, diameter 2cm, placed at waist height. Instructions emphasized to 

accomplish the task as fast as possible and keeping the hand opening range as 

close as possible to the diameter <strong>of</strong> the object; the distance dHand was then 

analysed in relation to the instruction given; 

2. ability in proportional control: the subject was asked to modulate the velocity 

<strong>of</strong> the elbow during the unloaded and loaded flexion-extension trials. An 

<strong>analysis</strong> <strong>of</strong> the EMG signals generated by the patient was carried out, verifying 

his strategies for proportional control; 

3. ability in CoCo: to assess the control <strong>of</strong> (switching between) different joints, 

sequential testing <strong>of</strong> alternating hand and wrist activations was performed by 

the patient. Assuming as starting configuration the wrist selected in full 

pronation, the instructed sequence consisted <strong>of</strong>: hand open-close, wrist 180° 

supination (first phase, P1), hand openclose wrist 180° pronation (phase 2, P2). 

The exercise required to correctly run through the all sequence <strong>of</strong> motors, not 

activating the elbow in between. P1 and P2 were repeated 6 and 5 times, 

respectively. For each repetition, the number <strong>of</strong> failed CoCos was quantified as 

long as the time required to complete the repetition. 

Questionnaire for prosthesis qualitative assessment 

The patient filled out a prosthesis- specific questionnaire [8] on his experience 

with this prosthesis and he compared it with the previous one. The 

questionnaire addressed issues related to performances, usability (use in ADLs, 

control, battery, failures), comfort and appearance, which had to be classified as 

positive or negative features <strong>of</strong> the prosthesis. 

Results 

143

Considering the elbow performance test, for each load the average flexion 

(extension) pattern was selected among those obtained and reported in figure 

5a,b, along with the confidence intervals for time duration and maximum 

flexion (extension) if exceeding 0.3s and 3°. Considering the flexion trials, the 

vertical confidence interval points out the highest, the lowest and the median 

maximal flexion among the repetitions. The horizontal, instead, reports the 

longest, the shortest and the median duration to reach the maximal flexion. 

Similarly for extension. The maximum elbow flexion ranged from 139° (0 Kg), 

to 125° (0.5Kg), to 118° (3 Kg), while the minimum extension from 7° (0.5Kg) 

to 16° (3Kg). The maximal and mean velocities reached in flexion were 266°/s 

and 164°/s, when no load was applied to the hand; these velocities decreased to 

59°/s and 36°/s for 0.5Kg and finally to 41°/s and 24°/s for 3Kg. The maximal 

and mean velocities in extension were 387°/s and 179°/s respectively, for 0Kg; 

these velocities decreased to 83°/s and 36°/s for 3Kg. The maximum latency 

measured for each weight was 0.12s, 0.29s, 0.18s, 0.57s, 0.54s, 0.41s in flexion 

and 0.06s, 0.05s, 0.05s, 0.48s, 0.66s, 0.58s in extension. Table 1 summarizes 

the results concerning the reliability <strong>of</strong> the elbow: the overall number <strong>of</strong> 

flexions (extensions) acquired is reported, together with the number <strong>of</strong> those 

failed. Moreover, these last are further characterized by pointing out the 

number <strong>of</strong> EMG activations required prior to obtaining the <strong>motion</strong> (Fig. 6a,b). 

Table 1- Results from the reliability assessment: the overall number <strong>of</strong> flexions (extensions) 

acquired is reported (columns 2-3), together with the number <strong>of</strong> the failed activations (columns 4- 

5). Each failed activation is further characterized by pointing out the number EMG activations 

required prior to obtaining the <strong>motion</strong> (columns 6-7). 

144

Fig. 5a, b - Elbow flexion (a) and extension (b) average selected patterns for the difference loads 

tested. For each load the median maximum flexion and minimum extension over the repetitions is 

reported in reverse colour (black dot for grey lines and grey dot for black lines). For each load the 

confidence interval for duration (angle) was reported only if the difference between the maximum 

and minimum duration (angle) among the repetition 

exceeded 0.3s (3°). Confidence intervals are reported in reverse colour too. 

Fig. 6a, b - Examples <strong>of</strong> failed activations, both in elbow flexion and extension. 

During the elbow performance test, the shoulder girdle EL-DE and PR-RE 

remained very limited, ranging between 2.5° and 5°. Similarly the head lateral 

flexion always remained within 5°. Shoulder flexion-extension tended to follow 

the elbow movements, swinging <strong>of</strong> about 15° – 20° mostly due to inertial 

effects during elbow extension. Considering the fine hand control exercise, the 

subject was always able to appropriately follow the recommendations. Over 4 

repetitions <strong>of</strong> the sequence reported, the subject overshot the diameter <strong>of</strong> at 

most 1 cm, being dHand (i.e. the inter-marker distance), 4 cm at hand closed, 6 

cm with the hand closed on the object and 10cm fully opened. The <strong>analysis</strong> for 

ability in proportional control during the elbow flexion-extension exercises, 

highlighted the ability <strong>of</strong> the subject to consciously modulate the EMG signals 

when instructed. As reported in figure 7a,b M.G. was able to perform both a 

145

linear-like and a step-like proportional control. For what concerns the CoCo 

ability assessment, figure 8a reports a correct sequence <strong>of</strong> the exercise: the 

active motor becomes the hand and an opening and closing is performed; then 

two CoCos bring the patient to the wrist through the elbow; a pronation ends 

the repetition. 

Figure 8b reports instead a typical sequence with recurrent failed CoCos. Over 

the 11 repetitions (considering both P1 and P2), only 1 was performed without 

errors. On average, the number <strong>of</strong> failed CoCos per repetition was 3.45 ± 2.87 

(mean ± 1 std), which lengthens the time required to complete it from 7.87s <strong>of</strong> 

the correct to 11.28 ± 2.77 for those with errors. On the all, for completing the 

11 repetitions M.G. took almost 2 minutes, due to 38 failed CoCos. Results 

from the questionnaire are reported in table 2. In general, conclusions were 

positive towards the performance and usability <strong>of</strong> the Dynamic Arm prosthesis. 

Its velocity, smooth (proportional) control and liftable loads were identified as 

the most beneficial aspects. The cosmetics, the weight and the possibility <strong>of</strong> an 

independent control <strong>of</strong> the elbow through dedicated myo-sites were the aspects 

to improve. 

Discussion 

The assessment method presented here was intended to help answering in an 

objective way two interconnected problems in myoelectric prostheses 

<strong>development</strong> and use: on one hand, the early testing <strong>of</strong> innovative prostheses in 

real-life conditions providing manufacturers with quantitative information 

regarding the prosthesis performance and reliability; on the other, the 

increasing need for quantitative assessment <strong>of</strong> the ability in myoelectric control, 

thus facilitating the cooperative work <strong>of</strong> CPOs in tuning the prosthesis and <strong>of</strong> 

OTs in focus training. In authors intentions, the proposed quantitative 

techniques are intended to integrate and sustain qualitative feedback about 

patient‘s feelings regarding pros and cons <strong>of</strong> the prosthesis, as well as clinical 

rating scales evaluations: e.g the authors strongly believe in the combined use 

<strong>of</strong> the control exercises proposed here with the promising ACMC scale [7]. For 

what concerns the elbow performances, the maximum flexion without loads is 

very close to that <strong>of</strong> able-bodied subjects [10]. When lifting loads, the flexion 

decreases proportionally, reasonably for safety reasons, while still adequate for 

a wide range <strong>of</strong> ADLs [2]. This is consistent with M.G. opinion. The velocity in 

flexion (with no load) and extension (up to 1Kg), even though reduced to 84% 

146

<strong>of</strong> the full performance, is almost that <strong>of</strong> able-bodied subject. Similarly to 

maximum flexion, with heavier loads velocity decreases, mostly for patient‘s 

safety. This is consistent with M.G. opinion about elbow performances. The 

confidence intervals reported stress the consistency <strong>of</strong> the elbow performances. 

Latencies between EMG activation and elbow <strong>motion</strong> suggest that with loads 

up to 1Kg the activation is almost immediate, while in the worst case lifting 

heavier loads the patient has to wait 0.7s before seeing the elbow move. 

Although the elbow would have been able to lift heavier loads, no tests were 

performed over 3 kg. This in fact was felt by the patient as the maximum 

weight affordable, due to high pressure in the socket and for personal safety 

during ex tension. Reliability <strong>analysis</strong> showed that failed activations appear 

both in flexion and in extension, with the exception <strong>of</strong> the unloaded condition. 

In flexions only they tend to increase with the load. The results obtained were 

reported to Otto Bock, which found the cause <strong>of</strong> the problem in the calibration 

<strong>of</strong> the mechanism responsible for the unlocking <strong>of</strong> the prosthesis when loaded 

(the safety mechanism was activated prematurely). The calibration was 

therefore modified before releasing the version <strong>of</strong> the elbow currently on the 

market. Results indicated a good control <strong>of</strong> the shoulder, shoulder girdle and 

hand during elbow tasks. No additional gross body movements are elicited nor 

needed to complete the task with increasing loads. This may partially be due to 

the use <strong>of</strong> biceps and triceps myo-sites for the control <strong>of</strong> the elbow in 

comparison with the results from trapezius and deltoid control in a patient 

(initials G. F.) with the same elbow but an amputation at a higher level, tested 

previously [5]. G. F., in fact, exhibited a shoulder <strong>motion</strong> ranging from 11° to 

8°, both for EL-DE and PR-RE, and a lateral flexion <strong>of</strong> the head from 8° to 12° 

toward the shoulder. In that case, a focused training and prosthesis tuning were 

considered. Results from the hand fine use and EMG modulation tests, proved 

the ability <strong>of</strong> the patient in exploiting the proportional control, his confident use 

<strong>of</strong> the terminal device and good prosthesis-patient interface. The results from 

the ability in the CoCo exercise, instead, brought in evidence a lack in patient‘s 

aptitude in switching between the prosthesis joints. Explanations for the poor 

performance can be theoretically sought for both in prosthesis set-up and in 

subject training. From the prosthesis side, it is known that a saturation <strong>of</strong> one <strong>of</strong> 

the channel can lead it to overcome the HIGH threshold (1.5 V) within 80 ms 

from crossing the ON threshold: this inhibits the recognition <strong>of</strong> CoCos. 

The <strong>analysis</strong> <strong>of</strong> the exercises excluded this situation: CoCos were ineffective 

due to M. G.‘s impossibility to raise both signals over the CoCo threshold or to 

raise both them over the CoCo threshold within 80ms. Results suggest therefore 

147

a focused re-training. Results appear consistent with M.G. desire to have 

independent EMG electrodes for controlling the elbow. An attempt to use a 

linear transducer in the harness for this purpose was un successful, due to the 

limitation it imposes on M.G. shoulder mobility. 

Conclusions 

A quantitative method to help assessing myoelectric prosthesis performance 

and reliability as well as the ability <strong>of</strong> patients in myoelectric control was 

proposed. The method was applied to test the premarket Otto Bock Dynamic 

Arm elbow fitted on a long-wearer trans-humeral amputee. The prosthesis 

performances appeared to be close to those <strong>of</strong> able-bodied subjects as long as 

amputee safety reasons do not limit them. Occasional unlocking <strong>of</strong> the elbow 

was reported to Otto Bock which took the observation into account for the 

release <strong>of</strong> the version currently on the market. The exercises for control 

assessment showed the ability <strong>of</strong> the patient to take advantage <strong>of</strong> the 

proportional control while highlighting difficulties in generating valid CoCos. 

In authors‘ opinion the proposed clinical/technological assessment can be 

valuable for CPO, manufacturers, clinicians, as well as for the patient himself. 

Prosthesis behaviour can be analysed for <strong>development</strong> purposes, highlighting 

inherent prosthesis problems. CPO and OT can benefit from these data by 

outlining problems due to the prosthesis-patient interface, to evaluate where 

interaction and performance <strong>of</strong> the patient and the prosthesis can be improved 

by individual adaptation <strong>of</strong> the prosthesis set-up or focused training 

interventions. To guarantee a wide applicability <strong>of</strong> the proposed patient 

quantitative assessment, current effort are focused on the substitution <strong>of</strong> the 

expensive stereophotogrammetric system for <strong>motion</strong> <strong>analysis</strong> with cheap 

sensors <strong>based</strong> on accelero meters and gyroscopes and on a simplification <strong>of</strong> the 

interconnections required for recording the EMG control signals: to this regards 

the collaboration with prosthesis producers will be strategic. 

Acknowledgements 

This work was partially supported by Regione Emilia-Romagna, PRRIITT 

Misura 3.4 Azione A in the framework <strong>of</strong> the StartER Project. 

148


1. Anglin, C. and Wyss, U. P.: Review <strong>of</strong> arm <strong>motion</strong> analyses. Proc Inst Mech 

Eng [H] 214 (2000), 541-555 

2. Buckley, M. A., Yardley, A., Johnson, G. R., and Carus, D. A.: Dynamics <strong>of</strong> 

the upper limb during performance <strong>of</strong> the tasks <strong>of</strong> everyday living – a review <strong>of</strong> 

the current knowledge base. Proc Inst Mech Eng [H ] 210 (1996), 241- 247 

3. Cappozzo, A., Della Croce, U., Leardini, A., and Chiari, L.: Human 

movement <strong>analysis</strong> using stereophotogrammetry. Part 1: theoretical 

background. Gait Posture (2004) XXX 

4. Cutti, A. G.: Motion <strong>analysis</strong> assessment <strong>of</strong> upper extremity motor ability. 

Ph.D. Thesis, University <strong>of</strong> Bologna (2006), 1-264 

5. Cutti, A. G., Davalli, A., Gazzotti, V., and Ninu, A.: Performance evaluation 

<strong>of</strong> the new Otto Bock ―Dynamic Arm‖ by means <strong>of</strong> biomechanical modelling. 

Proc. MEC ‗05, University <strong>of</strong> New Brunswick (2005), 197-201 

6. Cutti, A. G., Paolini, G., Troncossi, M., Cappello, A., and Davalli, A.: S<strong>of</strong>t 

tissue artefact assessment in humeral axial rotation. Gait Posture 21 (2005), 

341-349 

7. Hermansson, L. M., Fisher, A. G., Bernspang, B., and Eliasson, A.: 

Assessment <strong>of</strong> capacity for myoelectric control: a new Rasch-built measure <strong>of</strong> 

prosthetic hand control. J Rehabil Med 37 (2005), 166-171 

8. Janssens, K. and Cutti, A. G.: Prosthesis-specific questionnaire for the Otto 

Bock Dynamic Arm. (2006), 

http://www.inail.it/medicinaeriabilitazione/centroprotesi/pubblicazioni/doc/tec 

nica.htm 

9. Koerhuis, C. L., Winters, J. C., van der Helm, F. C., and H<strong>of</strong>, A. L.: Neck 

mobility measurement by means <strong>of</strong> the ‗Flock <strong>of</strong> Birds‘ electromagnetic 

tracking system. Clin Biomech (Bristol , Avon ) 18 (2003), 14-18 

149

10. Nordin, M. and Frankel, V. H.: Basic Biomechanics <strong>of</strong> the Musculo skeletal 

System. 2 (1989), Lea & Febiger, Philadelphia, London 

11. Prigge, P.: New concepts in external powered arm components. Proc. MEC 

‗05, 

University <strong>of</strong> New Brunswick (2005), 227-230 

12. Veeger, D. J., Yu, B., and An, K. N.: Orientation <strong>of</strong> axes in the elbow and 

forearm for biomechanical modeling. Proc <strong>of</strong> the 1st Conf <strong>of</strong> the ISG (1997) 

13. WHO: International Classification <strong>of</strong> functioning, disability and heath. 

(2001) 

14. Woltring, H. J.: Data processing and error <strong>analysis</strong>. In: Biomechanics <strong>of</strong> 

Human Movement, Ed Cappozzo A., Berme N. (1990), 203-237 

15. Wright, F. W.: Measurement <strong>of</strong> functional outcome with individuals who 

use upper extremity prosthesic devices: current and future directions. J Prosthet 

Orthot 18 (2006), 46-56 

150

2.3 DEVELOPMENT OF THE END-USER CLINICAL 

SOFTWARE FOR THE UPPER-EXTREMITY 

PROTOCOLS BASED ON STEREOPHOTOGRAMMETRY 

2.3.1 UpLiFE - Upper Limb Functional Evaluation Toolbox 

Figure 1 – UpLiFE toolbox GUI 

2.3.1.1 Design specifications 

UpLiFE (Upper Limb Functional Evaluation) Toolbox is a graphical user 

interface developed through Matlab allowing implementing the <strong>protocols</strong> <strong>based</strong> 

on the Vicon system for the upper limb 3D Kinematics on patients with 

shoulder pathologies, described in this Chapter. The s<strong>of</strong>tware is basically 

formed by 5 main components: 

1. Estimation <strong>of</strong> the anatomical landmarks position through palpation and 

estimation <strong>of</strong> the gleno-humeral center. 

2. Optimization <strong>of</strong> the clusters <strong>of</strong> markers and application <strong>of</strong> the CAST 

technique for each body segment. 

151

Girdle EL-DE (deg) 

3. 3D joint kinematics calculation following the coordinate systems 

definitions described in this Chapter. 

4. Creation <strong>of</strong> inter-joint coordination plots and bar plots (Figure 2-3) 

indicating the range <strong>of</strong> <strong>motion</strong> <strong>of</strong> each joint <strong>of</strong> interest, with the 

possibility to compare different sessions <strong>of</strong> measurement. 

45 

40 

35 

30 

Girdle EL-DE Vs GirdleHum F-E 

Arto Patologico Sessione 1 

Arto Patologico Sessione 2 

Arto Sano 

25 

20 

15 

10 

5 

0 

-5 

-20 0 20 40 60 80 100 120 140 

GirdleHum F-E (deg) 

Figure 2 – Outcome <strong>of</strong> UpLiFE Toolbox showing the comparison between the girdle-humeral 

rhythm in three different sessions <strong>of</strong> measurement 

152

degrees 

160 

140 

120 

117 

127 

150 

100 

80 

60 

40 

20 

43.7 

44.4 

120 

32.4 

34.5 

133 

30.9 

33.2 

172 

0 

-0.717 

-2.23 

-2.05 

-5.41 

-2.29 

-20 

-21.7 

-40 

Girdle EL-DE GirdleHum F-E Girdle EL-DE GirdleHum F-E Girdle EL-DE GirdleHum F-E 

Figure 3 – Outcome <strong>of</strong> UpLiFE Toolbox showing the comparison between the range <strong>of</strong> <strong>motion</strong>s <strong>of</strong> 

the angle components involved in the girdle-humeral rhythm in three different sessions <strong>of</strong> 

measurement 

UpLiFE Toolbox was created in order to reduce the time required for executing 

the upper limb <strong>protocols</strong> <strong>based</strong> on stereophotogrammetry developed at INAIL 

Prostheses Centre, in the routine application <strong>of</strong> the protocol itself and it as a 

valid tool for the validation studies. 

Particular attention was made to create the GUI in order to allow the user to 

easily go back into the steps <strong>of</strong> the data processing, save different 

configurations applied during the steps. 

Moreover, the interface can be easily enhanced and new coordinate 

systems/body segments definition can be easily created (e.g. when new body 

segments like the clavicle or the hand have to be considered). 

These features make UpLiFE a useful tool both for research and the routinely 

use <strong>of</strong> the <strong>protocols</strong>. 

153

2.3.1.2 UpLiFE Toolbox Tutorial 

1. exporting ASCII file from Vicon Nexus / Workstation 

2. creating the template folders 

3. positioning <strong>of</strong> the ASCII files to be used with UpLiFE interface: 

3.1 calibration files inside <strong>of</strong> ―Calibrations‖ folder 

3.2 static trial inside <strong>of</strong> calibration folder and task folder, if you want to 

compute kinematics <strong>of</strong> the static trial 

3.3 dynamic trials inside <strong>of</strong> task folder 

4. run mainGUI_UpLiFE 

4.1 select main directory 

4.2 anatomical Landmark calibration 

4.2.1 select, if you need, the stick used 

4.2.2 select, if you need, the names <strong>of</strong> the stick markers 

4.2.3 open calib list, answer yes if the update is needed 

4.2.4 open AL list 

4.2.5 tile vertically the two lists and change them so that each AL in the AL list 

has the corresponding calibration file, in the same row <strong>of</strong> the calib list 

4.2.6 save current state through the save button, if you prefer 

4.2.7 start CAST for getting the third point <strong>of</strong> the stick for each calibration file 


4.2.9 Estimation <strong>of</strong> GH / preliminary operations 

Scenario 1: the calibration file containing AC also includes IJ,PX,C7,T8 , 

required for the estimation <strong>of</strong> GH. In this case you can proceed to the next step. 

Scenario 2: the calibration file containing AC does not include one <strong>of</strong> the 

landmarks IJ,PX,C7,T8, all required for the estimation <strong>of</strong> GH. In this case you 

have to: 

a) press the ―copy files‖ button. The current calibration files will be copied in 

the ―Task‖ folder, ready to be used as dynamic trials, in order to apply 

solidification treatment. 

154

) go to the solidification panel and solidify the calibration file containing AC 

using the missing landmarks among IJ,PX,C7,T8 . For doing this, please follow 

the instructions provided in section 4.3. 

c) in the solidification panel, press the ―move files‖ button. In this way, the 

solidified files will be moved back to the ―calibration‖ folder and they will be 

now ready for the estimation <strong>of</strong> GH 

4.2.10 Estimation <strong>of</strong> GH / execution 

a) First select the side <strong>of</strong> the subject, right or left. 

b) One by one, use the browse buttons below ―EL file‖, ―EM file‖ and 

―AC file‖ to select the corresponding calibration file. 

c) Finally, press the ―start GH computation‖ button. The new estimation 

<strong>of</strong> GH will be saved in the calibration file containing AC. This must be taken 

into account for the next step <strong>of</strong> solidification, when the calibration file for GH 

must be included in the segment description. 

d) Repeat this procedure for both the sides, if necessary. 

4.3 Solidification 

4.3.1 By the ―load‖ button you can get the pre-defined configuration for right, 

left, or both limbs solidification scenario. Hereinafter, by the ―Save as…‖ 

button you can always save the current state <strong>of</strong> the solidification configuration. 

Note that this can be extremely useful for future computation <strong>of</strong> the same data, 

for correcting errors, or to try new methods <strong>of</strong> solidification. 

4.3.2 open segment list and comment the body segments you are not interested 

in. 

4.3.3 open task list. This will show all the tasks available in the ―Task‖ folder. 

Delete from the list the files that you are not interested in. 

4.3.4 In the ―segment descriptors‖ panel, first edit each body segment you are 

interested in. For a description <strong>of</strong> the descriptor, please refer to the 

APPENDIX. 

4.3.5 Press the ―static task‖ button and use the ―calib wizard‖ button. By doing 

this, the lines referring to the calibration files in the segment descriptors will be 

automatically filled, being this correspondence inside <strong>of</strong> the file list filled in 

section 4.2 . The wizard is active when the ―calib wizard‖ toggle button is 

pressed. When it is not pressed, you must fill the last lines <strong>of</strong> the segment 

descriptors by yourself. See APPENDIX for information about this. 

4.4 Kinematic <strong>analysis</strong> 

155

4.4.1 From the kinematic <strong>analysis</strong> panel, highlight the necessary joints. 

4.4.2 Press the ―See available tasks‖ button to see the list <strong>of</strong> solidified files 

available. 

4.4.3 Press the ―Open config file‖ button to visualize the configuration file 

4.4.4. Copy the wanted files to be computed from the list in 4.4.2 to the list in 

4.4.3 . 

4.4.5 Fill the part regarding the alignment files to be considered and the part 

regarding the files to be used for the estimation <strong>of</strong> the MHA. 

4.4.6 Press the start button. The code will ask you if the subject is an amputee 

and if the MHA algorithm must be applied. 

156

157

CHAPTER 3 


LOWER-EXTREMITY THROUGH 

STEREOPHOTOGRAMMETRIC SYSTEMS 

ABSTRACT 



KINEMATICS OF TRANSFEMORAL AMPUTEES 


KINETICS OF TRANSFEMORAL AND TRANSTIBIAL AMPUTEES 


3.2 DEVELOPMENT OF THE END-USER CLINICAL SOFTWARE FOR THE LOWER- 

EXTREMITY PROTOCOLS BASED ON STEREOPHOTOGRAMMETRY 

3.2.1 LOLIFE - LOWER LIMB FUNCTIONAL EVALUATION TOOLBOX 

ABSTRACT 

This chapter describes the <strong>development</strong> <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> on 

stereophotogrammetry specifically designed for the 3D kinematics and kinetics 

<strong>analysis</strong> on transfemoral amputees during gait. 

In particular the protocol was developed taking into account different kind <strong>of</strong> 

knee prostheses and some <strong>of</strong> the methodologies presented in Chapter 2. 

Moreover, a comparison between two different methods for estimating 3D joint 

moments and inertial parameters for specific knee prostheses are presented. 

As in Chapter 2, the <strong>development</strong> <strong>of</strong> the s<strong>of</strong>tware tools required for the 

application <strong>of</strong> the protocol is provided. 

158


3.1.1 Development <strong>of</strong> a protocol for the evaluation <strong>of</strong> lowerextremity 

kinematics <strong>of</strong> transfemoral amputees 

Extracted from 

3D JOINT MOMENTS IN TRANSFEMORAL AND 

TRANSTIBIAL AMPUTEES: WHEN IS THE "GROUND 

REACTION VECTOR TECHNIQUE" AN ALTERNATIVE 

TO INVERSE DYNAMICS 

Fantozzi S, Gar<strong>of</strong>alo P, Cutti AG, Stagni R 

Submitted to Gait & Posture 


Automatic procedures commonly adopted for studying lower limb kinematics 

(e.g. Plug In Gait [1] ) have been <strong>of</strong>ten used also for representing the 

kinematics and dynamics <strong>of</strong> the prosthetic limb, in particular for above-knee 

amputees [2,3], assuming the artificial limb as formed by human body 

segments. This assumption can be a limitation when comparing prosthetic 

limbs with unimpaired limbs, or when comparing different prosthetic limbs 

each other, due to the fact that the artificial limbs can have different geometries, 

different numbers <strong>of</strong> degrees <strong>of</strong> freedom at each joint, and different inertial 

characteristics. From the other side, inverse dynamics calculation is strictly 

dependent on the accuracy <strong>of</strong> both kinetic data and kinematic data. Therefore, 

each kind <strong>of</strong> prosthesis potentially requires the design <strong>of</strong> an accurate and 

specific protocol which takes into account these elements. 

Methods 

The protocol includes: 

-the definition <strong>of</strong> the kinematic model representing the prosthetic limb; 

159

-the definition <strong>of</strong> landmarks and coordinate systems for each segment <strong>of</strong> the 

prosthetic limb 

Definition <strong>of</strong> the kinematic model 

The prosthetic limb is formed by the residual thigh, the socket, the prosthetic 

shank and the prosthetic foot (Figure 1a,b). The relative movement between 

thigh and socket was considered negligible and thus residual thigh and socket 

will indicate herein a single segment. 

Figure 1 - Sagittal view <strong>of</strong> the prosthetic limb, formed by the residual thigh, the socket, the 

prosthetic shank ( a) C-Leg (Ottobock Healthcare, Germany), b) Power Knee (Ossur, Iceland) ) and 

the prosthetic foot. Markers and anatomical landmarks represented in figure are defined in Table 3 

Since the thigh forms both the (natural) hip joint and the knee joint (through the 

socket), to best define the axes <strong>of</strong> rotation <strong>of</strong> both joints, two different 

coordinate systems were defined for the thigh: a proximal thigh coordinate 

system (CS) being used for describing the hip joint and a distal thigh CS for the 

knee joint. Moreover, two different CSs were considered for the shank 

160

segment: a proximal shank <strong>based</strong> on specific landmarks <strong>of</strong> the prosthetic shank 

and a distal shank <strong>based</strong> only on specific landmarks common to the prosthetic 

shank and the prosthetic foot. 

Segment 

Axes definition 

Pelvis 

: medio-lateral 

: upward 

: antero-posterior 

Proximal Thigh 

: upward 



Proximal Shank 



: upward 

Distal Thigh 

during the static posture, using pth as technical frame 

Distal Shank 

upward 



: 

Foot (technical frame) 

: centroid 

Foot 

during the static posture, using ft as technical frame 

Table 1 - . Definition <strong>of</strong> the anatomical CSs for the prosthetic limb. All the axes definitions are 

showed for the right side. All the vectors are expressed in global reference frame (G). : cross 

product; R: 3x3 rotation matrix; : indicates the relative orientation <strong>of</strong> the B CS with respect to the A 

CS; : indicates that the vector must be normalized. 

Pelvis and proximal thigh were considered as the segments forming the hip 

joint, which was assumed as a ball & socket. Distal thigh and proximal shank 

161

formed the knee joint while the distal shank and the foot formed the ankle joint. 

The knee and the ankle joints were assumed as hinge joints, with one rotation 

(flexion-extension) occurring about a medio-lateral axis passing through the 

two main flexion-extension axis screws <strong>of</strong> the prosthesis (for the knee) and the 

medial and lateral screws connecting the prosthetic shank to the prosthetic foot 

(for the ankle). This representation is consistent to the degrees <strong>of</strong> freedom <strong>of</strong> 

the specific prosthetic devices (both for C-Leg and Power Knee prostheses in 

the first case study). 

Definition <strong>of</strong> landmarks and coordinate systems 

Specific landmarks were defined for the definition <strong>of</strong> the anatomical coordinate 

systems for the two different prosthetic knees (Figure 1a,b), taking into account 

their degrees <strong>of</strong> freedom, geometry and external features. The landmarks were 

reconstructed using CAST technique [4]. The definition <strong>of</strong> the anatomical CSs 

for the prosthetic side is reported in Table 1 and the definition <strong>of</strong> the anatomical 

landmarks adopted for the construction <strong>of</strong> the CSs is reported in Table 2. The 

terminology used for the axes forming the anatomical CS is the same adopted 

in [5]. 

Anatomical landmarks (prosthetic side) 

HJC 

KJC 

RPSIS 

RASIS 

LPSIS 

LASIS 

LS 

MS 

FT1 

FT2 

FT3 

FT4 

FF 

FL 

FM 

FP 

Hip Joint Center 

Prosthetic Knee Joint Center 

Right Posterior Superior Iliac Spine 

Right Anterior Superior Iliac Spine 

Left Posterior Superior Iliac Spine 

Left Anterior Superior Iliac Spine 

Lateral flexion-extension axis Screw 

Medial flexion-extension axis Screw 

First Foot marker 

Second Foot marker 

Third Foot marker 

Fourth Foot marker 

Foot Forward screw at the prosthetic ankle 

Foot Lateral screw at the prosthetic ankle 

Foot Medial screw at the prosthetic ankle 

Foot Posterior screw at the prosthetic ankle 

Table 2 - Definition <strong>of</strong> the anatomical landmarks (and markers) adopted for the construction <strong>of</strong> the 

CSs <strong>of</strong> the prosthetic limb, defined in Table A1. 

162

The definition <strong>of</strong> the distal thigh anatomical CS and the foot anatomical CS 

require a static calibration posture in which the subject is standing still with no 

flexion at the knee and ankle joint. This can be easily obtained asking the 

amputee to lift <strong>of</strong>f the ground the prosthetic side and to extend the hip. During 

the static posture, a constant orientation between the proximal shank and the 

proximal thigh (treated as a technical frame) is calculated, and, during each 

dynamic trial, the orientation <strong>of</strong> the distal thigh anatomical CS is updated 

sample by sample <strong>based</strong> on the orientation <strong>of</strong> the proximal thigh. This 

procedure is required because <strong>of</strong> the prosthesis and the residual thigh are 

connected through the socket. With this procedure, the joint angle representing 

the flexion-extension <strong>of</strong> the knee prosthesis is mathematically null in the static 

posture. 

Similarly, for the foot anatomical CS, during the static posture a constant 

orientation between the distal shank and the foot (treated as a technical frame) 

is calculated and during each dynamic trial the orientation <strong>of</strong> the foot 

anatomical CS is updated sample by sample <strong>based</strong> on the orientation <strong>of</strong> the 

distal shank. This procedure is required because no specific landmarks are 

available for the foot prosthesis and with this definition the joint angle 

representing the dorsi-plantarflexion <strong>of</strong> the ankle is mathematically null in the 

static posture. 

Hip, knee, and ankle joint angles were obtained, sample by sample, by 

decomposing the relative orientation <strong>of</strong> the anatomical CSs forming the joint 

with the Euler sequence ZX‘Y‘‘. 

163

3.1.2 Development <strong>of</strong> a protocol for the evaluation <strong>of</strong> lowerextremity 

kinetics <strong>of</strong> transfemoral and transtibial amputees 

INVERSE DYNAMICS Vs GROUND REACTION FORCE 

VECTOR METHODS: APPLICATION ON LOWER LIMB 

AMPUTEES 

Fantozzi S, Gar<strong>of</strong>alo P, Cutti AG, Stagni R, Davalli A 

Gait & Posture, vol. 30, suppl. 1, pp. 61-62 (October 2009) 


Lower limb joint moments are routinely used in human movement <strong>analysis</strong> to 

evaluate the deficit resulting from pathologies or the efficacy <strong>of</strong> treatments in 

terms <strong>of</strong> joint function. The calculation <strong>of</strong> three-dimensional (3D) internal joint 

moments is typically obtained by means <strong>of</strong> two approaches: the inverse 

dynamics (ID) and the projection <strong>of</strong> the ground reaction force vector (GRFV). 

The latter method, although simpler with respect to the former one, does not 

take into account the gravitational and inertial contributions on 3D joint 

moment. On the other hand, the first method has intrinsic potential errors such 

as the estimation <strong>of</strong> inertial parameters, <strong>of</strong> joint positions and orientation, the 

noise in kinematic data. The aim <strong>of</strong> the present study was to investigate, from a 

biomechanical point <strong>of</strong> view, the differences between ID and GRFV. In 

particular, subjects with different level <strong>of</strong> amputation are analyzed. 

Methods 

In the ID approach, gravitational, inertial, and ground reaction contribution on 

net 3D internal joint moment were identified. In this way, it was possible to 

analitically recognize the GRFV estimate in the ID calculations. Regarding the 

ID approach, the lower limb was represented as a chain <strong>of</strong> three rigid segments 

(foot, shank, thigh) and the ankle, knee and hip were modelled as spherical 

joints. Newton-Euler mechanics was applied to each segment starting from the 

feet. Angular accelerations and velocities were computed using finite 

164

differentiation. Inertial parameters were for the unimpaired limb were taken 

from Zatsiorsky after De Leva [6]. 

4 subjects (2 transfemoral and 2 transtibial amputees) (29 ± 4 years-old) were 

analysed during gait. The transfemoral amputees were analyzed when fitted 

with a C-Leg (Ottobock, Germany) reactive knee prosthesis. The transtibial 

amputees were fitted with a dynamic foot. Kinematics and kinetics data for the 

unimpaired limbs were acquired using a Vicon (Oxford Metrics) system and 

two Kistler force plates. The C.A.S.T. technique [4] was applied. For each 

subject, internal moments <strong>of</strong> ankle, knee and hip <strong>of</strong> the unimpaired limb were 

analyzed using the above mentioned methods. 

Results 

During the stance phase, the differences between the two methods were 

negligible in the ankle and knee moments, for each subject <strong>of</strong> the two groups <strong>of</strong> 

amputees. Evident differences between the two methods could be observed in 

the hip flexion-extension moment, for the transfemoral amputees. An example 

is shown in Figure 1. As GRFV does not allow to estimate the joint moment 

during the swing phase, no comparison can be performed. 

165

Figure 1 Example <strong>of</strong> ID moment and GRFV moment comparison for a transfemoral and a transtibial amputee. 

166


The use <strong>of</strong> the ID approach seems to be more effective than GRFV approach 

when analyzing the hip flexion-extension moment, where inertial and 

gravitational contributions become larger, in particular for transfemoral 

amputees. The use <strong>of</strong> the ID method could be also potentially useful in the 

impaired limb, when appropriate inertial parameters <strong>of</strong> the prosthetic devices 

are available, as different prostheses have different inertial properties with 

different joint moments during the swing phase. 

167


1. Davis R, Ounpuu S, Tyburski D, Gage J. A gait <strong>analysis</strong> data collection and 

reduction technique. Human Movement Science. 1991;10:575-587. 

2. Schmalz T, Blumentritt S, Jarasch R. Energy expenditure and biomechanical 

characteristics <strong>of</strong> lower limb amputee gait: : : The influence <strong>of</strong> prosthetic 

alignment and different prosthetic components. Gait & Posture. 

2002;16(3):255-263. 

3. Segal AD, Orendurff MS, Klute GK, et al. Kinematic and kinetic 

comparisons <strong>of</strong> transfemoral amputee gait using C-Leg and Mauch SNS 

prosthetic knees. J Rehabil Res Dev. 2006;43(7):857-870. 







2005;38(5):981-992. 

6. de Leva P. Adjustments to Zatsiorsky-Seluyanov's segment inertia 

parameters. J Biomech. 1996;29(9):1223-1230. 

168


SOFTWARE FOR THE LOWER-EXTREMITY 

PROTOCOLS BASED ON STEREOPHOTOGRAMMETRY 

3.2.1 LoLiFE - Lower Limb Functional Evaluation Toolbox 

Figure 1 - LoLiFE Toolbox GUI 

3.2.1.1 Design specifications 

LoLiFE (Lower Limb Functional Evaluation) Toolbox is a graphical user 

interface developed through Matlab allowing implementing the <strong>protocols</strong> <strong>based</strong> 

on the Vicon system for the lower limb 3D Kinematics and Kinetics on 

amputees during walking, described in this Chapter. The s<strong>of</strong>tware is basically 

formed by 5 main components: 

1. Estimation <strong>of</strong> the anatomical landmarks position through palpation and 

estimation <strong>of</strong> the hip joint center 

169

2. Optimization <strong>of</strong> the cluster <strong>of</strong> markers and application <strong>of</strong> the CAST 

technique for each body segment 

3. 3D joint kinematics calculation following the coordinate systems 

definitions described in this Chapter 

4. Creation <strong>of</strong> a standard clinical report including the 3D joint kinematics 

5. 3D joint kinetic calculation through inverse dynamics approach 

LoLiFE Toolbox was created in order to reduce the time required for executing 

the lower limb <strong>protocols</strong> <strong>based</strong> on stereophotogrammetry developed at INAIL 

Prostheses Centre, in the routine application <strong>of</strong> the protocol itself and as a valid 

tool for the validation studies. 

Currently, LoLiFE supports the execution <strong>of</strong> the protocol considering two 

different kind <strong>of</strong> prostheses, both for Kinematics and Kinetics calculations: C- 

Leg (Ottobock Healthcare, Germany) and Power Knee (Ossur, Iceland). 

Particular attention was made to create the GUI in order to allow the user to 

easily go back into the steps <strong>of</strong> the data processing, save different 

configurations applied during the steps. 

Moreover, the interface can be easily enhanced and new coordinate 

systems/body segments definition can be easily created (e.g. when new 

prostheses have to be considered). 

These features make LoLiFE a useful tool both for research and the routinely 

use <strong>of</strong> the <strong>protocols</strong>. 

170

3.2.1.2 LoLiFE Toolbox Tutorial 

1. exporting ASCII file from Vicon Nexus / Workstation 

2. creating the template folders 

3. positioning <strong>of</strong> the ASCII files to be used with LoLiFE interface: 

3.1 calibration files inside <strong>of</strong> ―Calibrations‖ folder 

3.2 static trial inside <strong>of</strong> calibration folder and task folder, if you want to 

compute kinematics <strong>of</strong> the static trial 

3.3 dynamic trials inside <strong>of</strong> task folder 

4. run mainGUI_LoLiFE 

4.1 select main directory 

4.2 anatomical Landmark calibration 

4.2.1 select, if you need, the stick used 

4.2.2 select, if you need, the names <strong>of</strong> the stick markers 

4.2.3 open calib list, answer yes if the update is needed 

4.2.4 open AL list 

4.2.5 tile vertically the two lists and change them so that each AL in the AL list 

has the corresponding calibration file, in the same row <strong>of</strong> the calib list 1 


4.2.7 start CAST for getting the third point <strong>of</strong> the stick for each calibration file 


4.2.9 Estimation <strong>of</strong> HJC / preliminary operations 

Scenario 1: the static calibration file to be adopted for the HJC estimation 

already includes LASIS, LPSIS, RASIS, RPSIS, required for the estimation <strong>of</strong> 

HJC through method from Bell. In this case you can proceed to the next step. 

Scenario 2: the static calibration file to be adopted for the HJC estimation does 

not include LASIS, LPSIS, RASIS, RPSIS, required for the estimation <strong>of</strong> HJC 

through method from Bell, but only the pelvis cluster. In this case you have to: 

1 Note that some <strong>of</strong> the predefined ALs has a ― * ― as suffix. This is when the ALS is involved in 

the computation <strong>of</strong> reference frames for the prosthetic side. 

Moreover, all the anatomical landmarks <strong>of</strong> the prosthetic side has ― A ― as prefix, despite <strong>of</strong> the 

amputee side. This will the convention adopted hereinafter. 

171

a) press the ―copy files‖ button. The current calibration files will be 

copied in the ―Task‖ folder, ready to be used as dynamic trials, in 

order to apply solidification treatment. 

b) go to the solidification panel and solidify the static calibration file 

containing the pelvis cluster applying the LASIS, LPSIS, RASIS, 

RPSIS anatomical landmarks. For doing this, please follow the 

instructions provided in section 4.3. 

c) in the solidification panel, press the ―move files‖ button. In this 

way, the solidified file will be moved back to the ―calibration‖ folder 

and they will be now ready for the estimation <strong>of</strong> HJC. 

4.2.10 Estimation <strong>of</strong> HJC / execution 

a) Firstly select the which is the amputee side, if present. 

Otherwise select ―none‖. 

b) Secondly select the method you want to adopt for the HJC 

estimation 2 

c) Use the browse button below ―HJC file‖ to select the 

corresponding static calibration file. 

d) Finally, press the ―start HJC computation‖ button. The new 

estimation <strong>of</strong> HJC will be saved in the static calibration file. This must 

be taken into account for the next step <strong>of</strong> solidification, when the 

calibration file for HJC must be included in the segment description. 

Note that for the amputee side the HJC will have the prefix ―A‖ and 

not ―R‖ or ―L‖ depending on the side. 

e) This procedure is automatically applied for both the sides. 

4.3 Solidification 

4.3.1 By the ―load‖ button you can get the pre-defined configuration for right, 

left, or both limbs solidification scenario, also considering different amputee 

side or prostheses. Hereinafter, by the ―Save as…‖ button you can always save 

the current state <strong>of</strong> the solidification configuration. Note that this can be 

extremely useful for future computation <strong>of</strong> the same data, for correcting errors, 

or to try new methods <strong>of</strong> solidification. 

4.3.2 open segment list and comment the body segments you are not interested 

in. 

2 Note that only the method from Harrington is still not implemented, being dependend on other 

information about the prosthetic side. 

172

4.3.3 open task list. This will show all the tasks available in the ―Task‖ folder. 

Delete from the list the files that you are not interested in. 

4.3.4 In the ―segment descriptors‖ panel, first edit each body segment you are 

interested in. 

4.3.5 Press the ―static task‖ button and use the ―calib wizard‖ button. By doing 

this, the lines referring to the calibration files in the segment descriptors will be 

automatically filled, being this correspondence inside <strong>of</strong> the file list filled in 

section 4.2 . The wizard is active when the ―calib wizard‖ toggle button is 

pressed. When it is not pressed, you must fill the last lines <strong>of</strong> the segment 

descriptors by yourself. 

4.4 Kinematic and dynamic <strong>analysis</strong> 

First choose one <strong>of</strong> the scenarios available through the ―Select scenario‖ popup 

menu. 

By selecting scenario 1 the code will assume that data from force plates is 

available and the corresponding .c3d file will be opened and information about 

gait events will be read during the Kinematics <strong>analysis</strong>. This scenario is 

therefore compatible with both the Kinematics and Dynamics <strong>analysis</strong>, 

including the generation <strong>of</strong> the standard clinical report. 

By selecting scenario 2, only the Kinematics <strong>analysis</strong> is possible, but not the 

generation <strong>of</strong> the standard clinical report. 

4.5 Kinematic <strong>analysis</strong> 

4.5.1 From the kinematic <strong>analysis</strong> panel, highlight the necessary joints. 

4.5.2 If force plates data is available, fill in the ―Analog freq (Hz)‖ box with the 

corresponding analog sampling frequency used during the acquisition 3 . 

4.5.3 Press the ―Open task list‖ button to see the list <strong>of</strong> solidified files available. 

4.5.4 Press the ―Open config file‖ button to visualize the configuration file 

4.5.5. Copy the wanted files to be computed from the list in 4.5.3 to the list in 

4.5.4 . 

4.5.6 Fill the part regarding the files to be used for the estimation <strong>of</strong> the MHA 4 . 

4.5.6 If amputee side is present you should first select the orthogonal static 

required for prosthesis alignment, by pressing the ―Browse‖ button. 

4.5.7 Press the ―Start Kine‖ button. The code will try to compute all the joints 

you selected. If some problem occur, please check if the correct ―Amputee 

3 Note that this is essential for correctly manage kinematic and kinetic data altogether. 

4 This feature is currently not available. The config file contains this part for easy future 

implementations. 

173

side‖ was selected through the interface. If not selected yet, the code will ask 

you to browse the orthogonal static calibration file. 

The code will run the kinematic <strong>analysis</strong>, asking you for further information 

like the method to be applied for the computation <strong>of</strong> the reference frames, the 

kind <strong>of</strong> prosthesis the subject is fitted with and the direction <strong>of</strong> walking 5 . The 

latter is necessary to understand which foot contact corresponds to which force 

plate 6 . 

4.5.8 If force plates data was available, click on ―Kine Report‖ for creating the 

standard clinical report considering all the available gait cycles previously 

computed. 

5 Dynamic <strong>analysis</strong> 

Please note that the dynamic <strong>analysis</strong> can be applied only limb by limb 7 . 

5. 1 If dynamic trials computed through scenario 1 are available, you can go on 

the ―Dynamics‖ panel and click on ―Open config‖. 

A new interface will appear, by which all the settings required for the inverse 

dynamics approach can be selected: 

- Height <strong>of</strong> the subject 

- Weight <strong>of</strong> the subject 

- Genre <strong>of</strong> the subject 

- Limb to be analyzed 

- Type <strong>of</strong> limb (prosthetic side or not) 

- Type <strong>of</strong> prosthesis (C-Leg and Power Knee are available) 

- Direction <strong>of</strong> walking 

- Human parameters (when selecting ―yes‖, human inertial 

parameters will be adopted despite <strong>of</strong> the type <strong>of</strong> limb. Selecting ―no‖, 

specific inertial parameters will be computed depending on the Type 

<strong>of</strong> prosthesis). 

- Moment normalization (this is only used for visualization) 

- Table <strong>of</strong> parameters (the anthropometric table considered as a 

reference) 8 

5 Note that the direction <strong>of</strong> walking is conventionally considered like ―forward‖ when moving along 

the positive y axis <strong>of</strong> the Vicon global frame, or ―backward‖ when moving the other way around. 

The direction <strong>of</strong> the y axis is only dependent on how the static calibration <strong>of</strong> the Vicon system was 

performed, i.e. where the L-Frame was positioned. 

6 Two force plates are assumed. 

7 Both the limbs can be computed in future versions simply running twice the inverse dynamic 

code. 

174

The task selection panel already visualizes the trials available in the 

\TaskAnalysis folder. In order to run the inverse dynamics code, the static task 

must also be selected, being necessary for the calculation <strong>of</strong> the subject specific 

body segment lengths. 

You can save and load your configurations by the save and load button 9 . 

5.2 Press the close button to close the interface. All the settings you applied 

will be maintained until you will open the small interface again. 

5.3 Press ―Start Dyn‖ to run the dynamic <strong>analysis</strong> on the previously selected 

trial. 3 figures will be created, one for each lower limb joint. 

8 Only the Zatsiorsky after De Leva is available. 

9 The configuration file will save all the settings except for which dynamic and static trial was 

selected. 

175

176

CHAPTER 4 


LOWER-EXTREMITY THROUGH 

INERTIAL AND MAGNETIC 

MEASUREMENT SYSTEMS 


4.1.1 OUTWALK PROTOCOL 



4.2.1 VALIDATION OF OUTWALK PROTOCOL ON BELOW-KNEE AMPUTEES 

4.2.2 EVALUATION OF ABOVE-KNEE AMPUTEES KINEMATICS DURING GAIT USING 

INERTIAL SENSORS 



PROTOCOLS BASED ON INERTIAL SENSORS 

4.3.1 DESIGN OF OUTWALK MANAGER AND MAIN FEATURES 

4.3.2 USE OF OUTWALK MANAGER IN CLINICAL SETTINGS 

4.3.3 OUTWALK MANAGER TUTORIAL 

ABSTRACT 

A complete description <strong>of</strong> ―Outwalk‖, a protocol <strong>based</strong> on inertial sensors 

specifically designed for the functional evaluation <strong>of</strong> cerebral palsy children 

and lower limb amputees during gait is presented. Outwalk was also validated 

on a population <strong>of</strong> below-knee and above-knee amputees. For the second group 

<strong>of</strong> amputees, in presence <strong>of</strong> magnetic disturbances, the method described in 

Chapter 6 was applied and validated. Finally, the design and a tutorial <strong>of</strong> an 

end-user clinical s<strong>of</strong>tware for the application and validation <strong>of</strong> Outwalk is 

presented. 

177


178

4.1.1 Outwalk protocol 

„OUTWALK‟: A PROTOCOL FOR CLINICAL GAIT 

ANALYSIS BASED ON INERTIAL & MAGNETIC 

SENSORS 

Cutti AG, Ferrari Al, Gar<strong>of</strong>alo P, Raggi M, Cappello A, Ferrari Ad 

Med Biol Eng Comput. 2010;48(1):17-25. 

Abstract 

A protocol named Outwalk was developed to easily measure on children with 

cerebral palsy and amputees the thorax-pelvis and lower-limb 3D kinematics 

during gait in free-living conditions, by means <strong>of</strong> an Inertial and Magnetic 

Measurement System (IMMS). Outwalk defines the anatomical/functional 

coordinate systems for each body segment through three steps: 1) positioning 

the Sensing Units (SUs) <strong>of</strong> the IMMS on the subjects‘ thorax, pelvis, thighs, 

shanks and feet, following simple rules; 2) computing the orientation <strong>of</strong> the 

mean flexion-extension axis <strong>of</strong> the knees; 3) measuring the SUs‘ orientation 

while the subject‘s body is oriented in a predefined posture, either upright or 

supine. If the supine posture is chosen, e.g. when spasticity does not allow to 

maintain the upright posture, hips and knees static flexion angles must be 

measured through a standard goniometer and input into the equations that 

define Outwalk anatomical coordinate systems. To test for the inter-rater 

measurement reliability <strong>of</strong> these angles, a study was carried out involving 9 

healthy children (7.9±2 year-old) and 2 physical therapists as raters. Results 

showed a RMS error <strong>of</strong> 1.4° and 1.8° and a negligible worst-case standard error 

<strong>of</strong> measurement <strong>of</strong> 2.0° and 2.5° for hip and knee angles, respectively. Results 

were thus smaller than those reported for the same measures when performed 

through an optoelectronic system with the CAST protocol, and support the 

beginning <strong>of</strong> clinical trials <strong>of</strong> Outwalk with children with cerebral palsy. 

179

Glossary 

CP: cerebral palsy 

CS: coordinate system 

DR: right drop-rise 

IMMS: inertial and magnetic measurement system 

h: hip static flexion angle during Outwalk calibration in supine posture 

k : knee static flexion angle during Outwalk calibration in supine posture 

PAT: posterior-anterior tilting 

RMS: root mean square 

SEM se : standard error <strong>of</strong> measurement considering systematic errors 

SEM nse : standard error <strong>of</strong> measurement not considering systematic errors 

SU: sensing unit <strong>of</strong> an IMMS 

TP: joint representing the movements <strong>of</strong> the Pelvis relative to the Thorax 

T1, T2: physical therapists involved in the reliability study <strong>of</strong> h and k 

180


Instrumental gait <strong>analysis</strong> has become a valuable tool in clinical practice [1] 

establishing its usefulness particularly for children with Cerebral Palsy (CP) 

[7], but also for amputees [15, 24]. Nevertheless, its use is still limited to very 

few medical centres, and its ability to monitor a patient‘s spontaneous and 

typical walking capacity, opposed to best performance, has still to be fully 

explored. In our opinion, the reasons for the present situation can be 

predominantly traced back to limitations in the measurement systems. In 

particular, optoelectronic systems are costly, hardly portable, and have a 

restricted field <strong>of</strong> view [3]. These features limit their use to dedicated 

laboratories and restrict the acquisition <strong>of</strong> a subject‘s gait to few strides per 

trial, in conditions which can be far from steady state. In addition, the 

acquisition <strong>of</strong> gait in an unfamiliar and artificial environment, such as that <strong>of</strong> a 

laboratory, can psychologically condition the subject, who will over-perform 

with respect to his/her every-day-life ability [13]. 

The recent availability <strong>of</strong> Inertial & Magnetic Measurement Systems (IMMSs) 

might open new perspectives for the measurement <strong>of</strong> the gait kinematics. 

IMMSs are commercially available, low-cost, and portable <strong>motion</strong> <strong>analysis</strong> 

systems. Thanks to these features, IMMSs might allow the user to execute and 

acquire the movement in laboratory-free settings, in a continuous modality, for 

long periods; therefore they might allow the user to collect a great number <strong>of</strong> 

consecutive gait cycles during spontaneous walking in daily life environments 

[16]. 

An IMMS consists <strong>of</strong> Sensing Units (SUs), which are lightweight boxes. Each 

SU integrates one 3D accelerometer, gyroscope, and magnetometer. The data 

supplied by these sensors are combined [21, 22] in order to measure the 3D 

orientation (but not the position) <strong>of</strong> the SU‘s Coordinate System (CS) with 

respect to a global, earth-<strong>based</strong> CS. Given this 3D orientation, an IMMS has the 

potential to estimate joint kinematics when: 1) a SU is attached to each bodysegment 

<strong>of</strong> interest; 2) at least one anatomical CS is defined for each bodysegment; 

and 3) the orientation <strong>of</strong> the anatomical CS is expressed in the CS <strong>of</strong> 

the SU. Joints kinematics can then be obtained from the relative orientation <strong>of</strong> 

contiguous anatomical CSs. The critical part <strong>of</strong> this process is the definition <strong>of</strong> 

the anatomical CSs. In fact, the lack <strong>of</strong> information regarding the position <strong>of</strong> 

the sensors implies that the anatomical CSs cannot be defined through the 

calibration <strong>of</strong> single anatomical landmarks (as recommended by the ISB [32]). 

181

Therefore different techniques must be conceived. 

Only a few studies investigated the use <strong>of</strong> IMMSs for gait kinematics 

measurement [18, 19]. O‘Donovan and co-workers [18] defined a protocol for 

3D inter-segment joint-angle measurement. However they specified their 

techniques just with respect to the ankle joint. In addition, the calibration steps 

require a rotation about the longitudinal axis <strong>of</strong> the whole body and a knee 

extension with minimal movement <strong>of</strong> the ankle, all tasks that may not be easily 

performed by certain populations <strong>of</strong> subjects, e.g. children with CP or 

amputees. 

Picerno and co-workers [19] defined a protocol to estimate lower-limb joint 

kinematics. The definition <strong>of</strong> the anatomical CSs was partially <strong>based</strong> on the 

external anatomical landmarks described in [6]. However, this protocol requires 

for each body segment to establish the orientation <strong>of</strong> a minimum <strong>of</strong> two nonparallel 

lines using multiple calibration tasks, involving multiple specialized 

devices at the expense <strong>of</strong> simplicity and experiment duration. 

In the effort <strong>of</strong> overcoming current limitations, the aim <strong>of</strong> this work was 

tw<strong>of</strong>old. 

First, it was intended to develop a new protocol, named ‗Outwalk‘, satisfying 

the following constraints: 1) suitable for IMMSs; 2) able to measure the 

kinematics <strong>of</strong> the TP (pelvis relative to the thorax), hip, knee and ankle joints; 

oriented to children with CP and to lower-limb amputees, and therefore <strong>based</strong> 

on 3) fast sensors mounting; 4) fast and comfortable calibration procedures; 5) 

not requiring any additional specialized device than the IMMS itself. 

Second, it was intended to test an essential requirement for Outwalk validity, 

i.e. to assess the inter-rater reliability <strong>of</strong> the goniometric measure <strong>of</strong> hip and 

knee static flexions in children laying supine on a mat. High reliability is 

searched for, as these measures are required as input to Outwalk when applied 

on subjects with irreducible knee flexion (e.g. in some forms <strong>of</strong> CP), laxity or 

deformities. Our hypothesis was that static hip and knee flexions can be 

measured with a precision not lower than that reported in [11] for hip and knee 

flexions measured with an optoelectronic system through the CAST protocol, 

assumed as clinical reference. Confirmation <strong>of</strong> this hypothesis would support 

the commencement <strong>of</strong> the clinical trial <strong>of</strong> Outwalk in CP children. 

182

2. METHODS 

2.1 DEVELOPMENT OF ‗OUTWALK‘ 

In developing Outwalk, the approach described in [8] for the upper-limb was 

used as reference. 

2.1.1 TARGET POPULATION 

Outwalk was designed to be suitable for above and below knee amputees and 

for children with CP. In particular, CP children can be both hemiplegic 

belonging to forms I- III <strong>of</strong> Winters et al. [30] and diplegic belonging to forms 

II- IV <strong>of</strong> Ferrari et al. [14]. 

2.1.2 REQUIREMENTS FOR THE MEASUREMENT SYSTEM 

Outwalk was conceived for a <strong>motion</strong> tracking system: 1) capable to measure in 

time the orientation <strong>of</strong> the local CSs <strong>of</strong> its SUs with respect to a global CS; 2) 

featuring a visible reference between the orientation <strong>of</strong> the local CS and the 

physical appearance <strong>of</strong> its SU (e.g. the box containing the electronics <strong>of</strong> the 

SU), with the smallest possible misalignment. 

The Xsens system (Xsens Technologies, NL), is an IMMS which satisfies these 

requirements (Fig. 1 – on-line material). As this was used in [12] for the 

validation <strong>of</strong> Outwalk, herein we will only explicitly refer to this IMMS. The 

Xsens consists <strong>of</strong> up to 10 SUs (called MTx) connected by-wire to a datalogger 

(Xbus Master), usually worn on the belt. The data-logger is connected 

via Bluetooth to a laptop which processes and stores the data collected. Each 

SU is hosted in a small box, weights 38g, and is 39x54x28mm. The local CS <strong>of</strong> 

the SU is aligned with the boundaries <strong>of</strong> the box with an error < 3° (Xsens 

Technical Manual). The orientation <strong>of</strong> each SU‘s CS with respect to an earth<strong>based</strong> 

global CS is provided as an output. 

2.1.3 DEFINITION OF THE REFERENCE KINEMATIC MODEL 

The definition <strong>of</strong> the kinematic model <strong>of</strong> Outwalk was achieved by following 

the indications provided in [17]. 

Thorax, pelvis, thigh, shank and foot were assumed as the rigid segments 

forming the TP, hip, knee and ankle joints. 

183

TP, hip and ankle joints were assumed as ball & sockets. The knee was 

assumed as a ‗loose‘ double hinge-joint, with one rotation (flexion–extension) 

occurring about a mediolateral axis fixed in the distal femur and the other 

rotation (internal-external) occurring about a longitudinal axis fixed in the tibia 

[23]. The double-hinge is defined ―loose‖ (using a term common for elbow 

endoprostheses), since it actually allows some ab-adduction. However, when 

the flexion-extension and internal-external rotations axes are correctly located, 

movement <strong>of</strong> the knee joint within a range <strong>of</strong> 5–90° <strong>of</strong> flexion can be almost 

entirely accounted for by simultaneous rotations about these two axes [23]. 

Following the standard Denavit-Hartenberg convention used in robotics [26], 

for each segment we defined as many CSs as the number <strong>of</strong> joints the segment 

forms, that is: one for the thorax and foot, and two for pelvis, thigh and shank. 

In the authors‘ opinion, this approach mitigates a limitation <strong>of</strong> the current ISB 

standard, in which a single set <strong>of</strong> orthogonal axes is defined for a segment, and 

used to describe rotations <strong>of</strong> different joints. This is for example the case <strong>of</strong> the 

thigh anatomical CS, in which the Y axis is considered as the hip internalexternal 

rotation axis, and Z the knee flexion-extension axis. Unfortunately, the 

orientation <strong>of</strong> Z is not directly controlled, as it derives from Y and X; as a 

consequence, it can be generally different from the (mean) axis <strong>of</strong> rotation <strong>of</strong> 

the knee, which should be preferably used instead [25]. 

As a general rule for naming, for each segment the CS describing the proximal 

joint is defined as the ―proximal CS‖, whereas the CS describing the distal joint 

is defined as the ―distal CS‖. Moreover, considering the right side <strong>of</strong> the body, 

in the segments‘ CSs the Y axis points cranially, Z laterally and X anteriorly 

[32]. For the left side CSs Y points caudally, Z medially and X posteriorly. By 

so doing, a clinical flexion, abduction or internal rotation performed with the 

left side <strong>of</strong> the body assumes the same positive or negative sign assumed when 

performed with the right side. In other words, the kinematic patterns <strong>of</strong> the left 

side can be directly plotted over those <strong>of</strong> the right side. 

The rotations describing the degrees <strong>of</strong> freedom <strong>of</strong> the joints were named: 

posterior-anterior tilting (PAT), right drop-rise (DR), right internal-external 

rotation (IE) for the TP joint; flexion-extension (FE), adduction-abduction 

(AA) and internal-external rotation (IE) for the hip; FE, varus-valgus (VV) and 

IE for the knee; dorsi-plantar flexion (DP), inversion-eversion (IV) and IE for 

the ankle. In each <strong>of</strong> these couples, the first rotation is expected to have a 

positive sign [32]. 

184

2.1.4 PROCEDURE TO MEASURE THE TP AND LOWER-LIMB 

KINEMATICS 

The procedure to measure the TP, hip, knee and ankle kinematics <strong>of</strong> a subject 

consists <strong>of</strong> the following 4 steps: 1) positioning the SUs on the subject‘ thorax, 

pelvis, thighs, shanks and feet; 2) computing the orientation <strong>of</strong> the mean FE 

axis <strong>of</strong> the knees in the thighs embedded CSs <strong>of</strong> the SUs; 3) defining 

anatomical/functional CSs for thorax, proximal pelvis, distal pelvis, proximal 

thighs, distal thighs, proximal shanks, distal shanks and feet, and expressing 

their orientation in the SU CS <strong>of</strong> the corresponding segment; 4) computing the 

joint-angles. These steps are described for the right leg only. 

2.1.4.1 Positioning the SUs 

One SU is positioned on each body-segment with double-sided tape, either over 

the skin or over elastic cuffs wrapped around the segments (Fig. 2). For the 

thorax, the SU is positioned over the flat portion <strong>of</strong> the sternum, with the Z axis 

<strong>of</strong> the SU pointing away from the body and X cranially. For the pelvis, the 

midline <strong>of</strong> the SU is aligned with the spine, and its X axis is oriented along the 

line linking the posterior superior iliac spines (PSIS), pointing toward the right 

PSIS. For the thigh, the SU is positioned laterally, within its median third. For 

the shank, the SU is positioned within the distal third, close to the lateral 

malleolus, with the X axis aligned with the long axis <strong>of</strong> the fibula. The Z axis 

<strong>of</strong> the SU points laterally in the body‘s frontal plane. For the foot, the base <strong>of</strong> 

the SU is positioned and oriented over the shoe in order to maximize its 

stability. In particular, it is recommended to place it over the midfoot, and 

ascertain that the orientation <strong>of</strong> the sensor is not affected by forefoot-midfoot 

relative <strong>motion</strong> during the third rocker. 

185

Figure 2 - Xsens SUs positioned over the body <strong>of</strong> a subject 

Figure 3 - Upright calibration posture. Red, green and blue arrows indicate the X, Y and Z axes <strong>of</strong> 

the Xsens local CS. 

186

2.1.4.2 Functional movements to compute the knee mean flexion-extension axis 

To define the anatomical CS for the distal thigh and to express its orientation in 

the SU CS <strong>of</strong> the segment, the direction <strong>of</strong> knee FE axis must be estimated first. 

The orientation <strong>of</strong> the SUs over thigh and shank is measured during a pure knee 

FE task. If the patient can perform the task autonomously, he is instructed to 

stand in the upright posture and, helped by an examiner in keeping the posture, 

to flex-extend the knee five times up to 70°. If the patient cannot stand in 

upright posture or autonomously execute the task, the task can be performed 

passively, with the subject laying in supine position, while a therapist executes 

the knee FE movement 5 times up to 70° <strong>of</strong> flexion. The direction <strong>of</strong> the knee 

FE axis is estimated using the functional method described in [31] and it is 

expressed in the CS <strong>of</strong> the SU <strong>of</strong> the thigh (V FLEX ). 

2.1.4.3 Static acquisition to complete the definition <strong>of</strong> the anatomical CSs 

The anatomical CS <strong>of</strong> the proximal pelvis is assumed to be coincident with the 

SU CS. 

To define the anatomical CSs for thorax, distal pelvis, proximal and distal 

thigh, proximal and distal shank, and foot and to express the orientation <strong>of</strong> 

these anatomical CSs in the SU CS <strong>of</strong> the corresponding segment, the 

orientation <strong>of</strong> the SUs‘ CSs is measured during a 5 seconds static trial. During 

this static trial the subject is asked to stand still in one <strong>of</strong> the following two 

postures: 

1) upright with the back straight, looking forward, knee centre aligned to the 

ASIS, and the line from the 2 nd metatarsal head to the calcaneus <strong>of</strong> the right 

foot parallel to the same line <strong>of</strong> the left foot (Fig. 3); 

2) in supine position on a mat, with the hip flexed h degrees and the knee flexed 

k degrees, knee centre aligned with the ASIS, feet in neutral position and 

parallel to each other as in posture 1 (Fig. 4). 

187

Figure 4 Supine calibration posture. Hips and knees can be flexed h° and k° degrees, respectively, 

in case <strong>of</strong> irreducible flexions <strong>of</strong> the subject. In this case, h and k must be measured through a static 

goniometer, and used to compute γ and δ; h, γ and δ are then input in the definition <strong>of</strong> the 

anatomical CSs (Table 2). If the subject wears AFOs, the zero-condition for the ankle joint-angles 

is the one assumed by the feet inside the AFO. THX: thorax; H: hip; K: knee; A: ankle. 

Posture 2 should be used with CP children and subjects with irreducible knee 

flexion, laxity or deformities (genu varum, valgum, recurvatum and flexum). 

Angles h and k will depend on the specific patient, and they will be 0° if the leg 

can be fully extended. To sustain the subject‘s legs, a typical therapeutic foam 

cylinder (or wedge) is recommended (Fig. 4). A second foam cylinder or wedge 

can be used to maintain the expected foot calibration posture. If the subject uses 

an ankle-foot orthosis (AFO), the foot calibration posture is that imposed by the 

AFO. In posture 2, if the SU on the pelvis touches the mat, than two separate 

mats should be used and kept as close as possible, but allowing a gap below the 

SU. The definitions <strong>of</strong> the anatomical CSs reported in Table 1 and Table 2 are 

then applied for the upright and supine posture, respectively. CSs were 

developed following the example <strong>of</strong> [8]; they have a constant orientation with 

respect to the SU‘s CS <strong>of</strong> the corresponding segment. Definitions in Table 2 are 

<strong>based</strong> on those <strong>of</strong> Table 1, but take into account that: 

1) hip and knee can be flexed h° and k°, and therefore the cranio-caudal axes <strong>of</strong> 

188

thigh and shank cannot be assumed to lie in the frontal plane. Their rotations 

out <strong>of</strong> the frontal plane must be compensated to properly define the CSs <strong>of</strong> 

proximal and distal thigh, and proximal shank. To apply the definitions in 

Table 2, therefore, the examiner will need to measure h and k through a static 

goniometer; 

2) when the calibration is executed in supine position, the Z axis <strong>of</strong> the SU on 

the thorax becomes almost parallel to the gravity line, and therefore the 

computation <strong>of</strong> a thorax medio-lateral direction from Z and gravity can be 

badly conditioned. 

It is worth noticing that the distal shank CS is assumed coincident with that <strong>of</strong> 

the proximal shank, since functional methods cannot still be reliably used for 

the estimation <strong>of</strong> the ankle axes <strong>of</strong> rotation [4]. 

189

Segment Axes Definition Anat. direction. 

Thorax (TH) 

Pelvis – Proximal (pPL) 

Pelvis - Distal (dPL) 

Thigh - Proximal (pTG) 

Thigh - Distal (dTG) 

Shank - Proximal (pSK) 

Shank - Distal (dSK) 

Y TH = SU-TH Z G / || . || 

Z TH = [0 0 1] Y TH / || . || 

X TH = Y TH Z TH / || . || 

X pPL = -[0 0 1] 

Y pPL = [0 1 0] 

Z pPL = [1 0 0] 

SU-PL R dPL = SU-PL R TH 

SU-TG R pTG = SU-PL R dPL 

Y = Y pSK; 

Z dTG = V FLEX / || . || 

X dTG = Y Z dTG / || . || 

Y dTG = Z dTG X dTG / || . || 

Y = Y TH; 

Z pSK = [0 0 1] 

X pSK = Y Z pSK / || . || 

Y pSK = Z pSK 

X pSK / || . || 

SU-SK R dSK = SU-SK R pSK 

cranial 

lateral 

anterior 

anterior 

cranial 

lateral 

lateral 

anterior 

cranial 

lateral 

anterior 

cranial 

Foot (FT) 

SU-FT R FT = SU-SK R dSK 

Table 1 – Table Definition <strong>of</strong> the anatomical/functional CSs (thorax, pelvis, and right leg), to be 

used when the subject stands in the upright posture during the static calibration. CSs which share 

the same gray code in the right column have the same orientation during the calibration posture. For 

each segment, all vectors are expressed in the CS <strong>of</strong> the SU positioned on the segment. 

Abbreviations. : cross product; R: 3x3 rotation matrix; A R B indicates the relative orientation <strong>of</strong> 

the B CS with respect to the A CS; || . ||: indicates that the vector must be normalized; Z G: Z axis <strong>of</strong> 

the global CS, assumed opposed to gravity; V FLEX: direction <strong>of</strong> the flexion-extension axis <strong>of</strong> the 

knee; SU-TH: SU on thorax; SU-PL: SU on pelvis; SU-TG: SU on thigh; SU-SK: SU on shank; 

SU-FT: SU on foot. 

190

Segment Axes Definition Anat. direction. 

Thorax (TH) 

Pelvis – Proximal (pPL) 

X TH = SU-TH Z G / || . || 

Z TH = X TH [1 0 0] / / || . || 

Y TH = Z TH X TH / / || . || 

X pPL = -[0 0 1] 

Y pPL = [0 1 0] 

Z pPL = [1 0 0] 

anterior 

lateral 

cranial 

anterior 

cranial 

lateral 

Pelvis - Distal (dPL) 

SU-PL R dPL = SU-PL R TH 

Thigh - Proximal (pTG) 

SU-TG R pTG = SU-TG R TH·R Z(+h°) 

Thigh - Distal (dTG) 

Shank - Proximal (pSK) 

Y = 2 nd column <strong>of</strong> SU-SK R pSK·R Z(+δ°); 

Z dTG = V FLEX / || . || 

X dTG = Y Z dTG / || . || 

Y dTG = Z dTG X dTG / || . || 

Y = 2 nd column <strong>of</strong> SU-SK R TH·R Z(-γ°); 

Z pSK = [0 0 1] 

X pSK = Y Z pSK / || . || 

Y pSK = Z pSK 

X pSK / || . || 

lateral 

anterior 

cranial 

lateral 

anterior 

cranial 

Shank - Distal (dSK) 

SU-SK R dSK = SU-SK R pSK 

Foot (FT) 

SU-FT R FT = SU-SK R dSK 

Table 2 Table 2 - Definition <strong>of</strong> the anatomical/functional CSs (thorax, pelvis, and right leg), to be 

used when the subject lies horizontally in supine position during the static calibration. CSs which 

share the same gray code in the right column have the same orientation during the calibration 

posture. For each segment, all vectors are expressed in the CS <strong>of</strong> the SU positioned on the segment. 

Angles h, k are the rotations in the sagittal plane <strong>of</strong> the hip and knee in the static posture, and have 

to be measured through static goniometry; δ = 180-k and γ = 180-h-k (see Fig. 4). Abbreviations. 

: cross product; R: 3x3 rotation matrix; A R B indicates the relative orientation <strong>of</strong> the B CS with 

respect to the A CS; R Z(+h°), R Z(+δ°), R Z(-γ°): indicate a constant rotation matrix around a Z axis 

<strong>of</strong> +h°, +δ° and –γ°; || . ||: indicates that the vector must be normalized; Z G: Z axis <strong>of</strong> the global CS, 

assumed opposed to gravity; V FLEX: direction <strong>of</strong> the flexion-extension axis <strong>of</strong> the knee; SU-TH: SU 

on thorax; SU-PL: SU on pelvis; SU-TG: SU on thigh; SU-SK: SU on shank; SU-FT: SU on foot.. 

191

2.1.4.4 Computation and interpretation <strong>of</strong> the joint-angles 

During data acquisition for a walking task, the orientation <strong>of</strong> the anatomical 

CSs is updated sample-by-sample <strong>based</strong> on the orientation <strong>of</strong> the SUs‘ CS. The 

TP, hip, knee, and ankle joint-angles are then obtained, sample-by-sample, by 

decomposing the relative orientation <strong>of</strong> the anatomical CSs forming the joint 

with the Euler sequence ZX‘Y‘‘ [32]. Finally, the sign <strong>of</strong> the Z‘ rotation for the 

knee is reversed in order to have a positive flexion, as declared in section 2.1.3 

and as expected in clinics [2]. 

It is worth noticing that, <strong>based</strong> on the definition for the distal thigh and 

proximal shank, the knee flexion in the static posture is 0° (since Z dTG lies in 

the plane normal to X dTG such as Y dTG , Y pSK ), as it is also the case for all hip 

and ankle angles. 

It has been widely demonstrated that the knee rotations other than the FE are 

strongly affected by the s<strong>of</strong>t-tissue artefact problem when measured through 

skin-mounted markers, and are therefore unreliable [20, 25, 27]. For this 

reason, in our future clinical applications these angles will not be generally 

analysed. 

2.2. INTER-RATER RELIABILITY OF THE GONIOMETRIC MEASURE 

OF HIP AND KNEE STATIC FLEXION 

High reliability in measuring h and k through a goniometer is required for the 

applicability <strong>of</strong> Outwalk when the supine calibration posture is used. Failure in 

this requirement would cast doubts about the <strong>of</strong>fset measured for hip and knee 

flexion-extension patterns during gait (see Table 2), which would substantially 

depend by the rater leading the measurements. 

Since the static posture will be typically used with CP children, a pre-clinicaltrial 

inter-rater reliability study was carried out by involving 9 healthy children 

and 2 raters. 

2.2.1 SUBJECTS AND RATERS 

Nine healthy children (7.9±2 year-old, 27.5±7 Kg, 129±12 cm tall; 6 males, 3 

females) were enrolled in the study after obtaining the informed consent from 

the parents. The study also involved two physical therapists (T1 and T2) as 

raters, with experience in the treatment <strong>of</strong> CP children. Before beginning the 

study, T1 and T2 practiced together in the measurement <strong>of</strong> h and k on two 

additional children who were not included in the study. 

192

2.2.2 EXPERIMENTAL SET-UP 

For the measurement <strong>of</strong> h and k, a standard plastic goniometer with two 

movable bars each <strong>of</strong> 20cm in length and 8cm in height was used. 

Given a child, this was asked to lie supine on a mat with a foam cylinder under 

the knees (as in Fig. 4). Then T1 and T2 independently measured h and k <strong>of</strong> one 

<strong>of</strong> the legs. Between the two measurements, the child was asked not to move. 

The child was then free to move for a few minutes and then T1 and T2 

independently measured h and k on the remaining side, as in the first 

measurement. The first side to be assessed was randomly selected, as well as 

the order <strong>of</strong> assessment by T1 and T2. 

To measure h, one bar <strong>of</strong> the goniometer was lain on the mat and the other 

oriented to be on the line between the greater trochanter and the lateral 

epicondyle <strong>of</strong> the femur. 

To measure k, one bar <strong>of</strong> the goniometer was oriented to be on the line between 

the greater trochanter and the lateral epicondyle <strong>of</strong> the femur, and the other to 

be on the line between the head <strong>of</strong> the fibula and the lateral malleolus. 

2.2.3 HYPOTHESIS AND DATA ANALYSIS 

To conclude that h and k have an inter-rater reliability adequate to begin the 

clinical trial <strong>of</strong> Outwalk, we tested the following hypothesis: 

H1: h and k reliability must be not lower than that reported in [11] for 

h and k measured with an optoelectronic system through the CAST protocol. 

To test this hypothesis, we considered the h and k angles measured for each leg 

as a set <strong>of</strong> independent observations, and we computed the following root mean 

squared errors (RMS), consistently with [11]: 

18 

i1 

( h 

h ) 

h 

RMS 

 

, 

18* 2 

2 

 

j1 

ij 

i 

2 

k 

RMS 

 

18 

i1 

2 

 

j1 

( k 

ij 

18* 2 

k ) 

i 

2 

where: 

i=1…18: number <strong>of</strong> legs examined; 

193

j=1…2: number <strong>of</strong> raters involved; 

hi1 h 

h i2 

i , mean hip flexion angle among T1 and T2 for leg i; 

2 

ki1 k 

k i2 

i , mean knee flexion angle among T1 and T2 for leg i. 

2 

Based on the results reported in [11], H1 is true if: 

h RMS 

k RMS 

5 

3. 7 

Following current recommendations about statistical parameters describing 

reliability [10], we also computed for h and k the Standard Error <strong>of</strong> 

Measurement (or ‗typical error‘), considering (SEM se ) and not considering 

(SEM nse ) the systematic error introduced by T1 and T2. Following [29], SEM se 

and SEM nse were obtained through a single-factor (i.e. raters) repeated 

measures ANOVA in the 1-way and 2-way model, respectively. 

3. RESULTS 

The original data measured by T1 and T2 on the 18 legs are reported in Table 3. 

hRMS 

and k RMS were found to be 1.4° and 1.8°, i.e. less than those reported 

in [11]. It was concluded that H1 was fully satisfied. For h, both SEM se and 

SEM nse were 2.0°, while for k they were 2.4° and 2.5°, respectively. 

194

LEG 

h° k° 

T1 T2 T1 T2 

1 142 138 120 124 

2 138 142 120 118 

3 140 144 120 115 

4 142 142 120 115 

5 152 148 120 114 

6 150 148 120 114 

7 145 141 110 111 

8 142 140 110 108 

9 145 145 110 108 

10 146 143 110 110 

11 142 139 114 119 

12 140 142 120 114 

13 144 144 110 110 

14 144 142 110 106 

15 134 134 108 108 

16 126 129 110 112 

17 140 137 116 115 

18 140 136 116 116 

Table 3 Values measured for angle h (hip static flexion) and k (knee static flexion) by rater T1 and 

T2, on the 18 legs examined (9 children). 

4. DISCUSSION 

Outwalk was developed to measure thorax-pelvis and lower-limb kinematics 

with IMMS, in clinical settings, and specifically in below/above knee amputees 

and CP children. In particular, the CP population described in section 2.1.1 was 

selected by considering the main goal <strong>of</strong> their rehabilitation treatments, that is 

the acquisition and preservation <strong>of</strong> gait, even for long distances. 

Since IMMS cannot measure the position <strong>of</strong> their SUs, Outwalk was not <strong>based</strong> 

on the calibration <strong>of</strong> single anatomical landmarks for the construction <strong>of</strong> the 

anatomical CSs. The protocol takes about 10min to complete from subject 

arrival, and it does not require any specialized device other than the SUs, in 

195

contrast to [19]. Moreover, the two calibration postures (upright or supine) 

were selected to be comfortable for our population <strong>of</strong> interest. In general, we 

expect to use the upright posture for amputees, while the supine posture for CP 

children. Amputees can in fact easily maintain the predefined upright posture, 

while this cannot be assumed for children with CP, who may present a flexed 

knee and [28] hip-knee-ankle flexion (scissor pattern [5]). Similar 

considerations apply for the functional estimation <strong>of</strong> the knee mean FE axis <strong>of</strong> 

rotation. In general, on CP children the movement will be passively executed 

by the therapist, while amputees will actively perform the movement, or 

maintain the upright posture while a rater passively flexes and extends the 

prosthetic knee (if any). 

Outwalk anatomical CSs were developed to best match the kinematic 

assumptions for the lower-limb joints. In particular, the definition <strong>of</strong> the distal 

knee CS presents an advantage with respect to the CS recommended by the ISB 

[32]. The advantage is that in Outwalk, the medio-lateral axis (Z) <strong>of</strong> the CS is 

defined along the mean FE axis <strong>of</strong> rotation <strong>of</strong> the knee, while in the ISB the 

femur Z axis is obtained as the last axis after the longitudinal and posterior axes 

are computed. This means that the medio-lateral axis <strong>of</strong> rotation <strong>of</strong> the knee in 

the ISB anatomical CS is not directly controlled, and its direction can be 

different from the inter-epicondilar axis [25]. Since a target population for 

Outwalk are the transfemoral amputees, the application <strong>of</strong> the ISB approach for 

the prosthetic knee would have been in explicit contradiction to the a-priori 

knowledge that the knee is a perfect hinge, with the axis <strong>of</strong> rotation oriented in 

a single direction. Outwalk, instead, respects this knowledge and takes it into 

account in the definition <strong>of</strong> the CS. 

As a counter-balance <strong>of</strong> Outwalk ease <strong>of</strong> use, the operation <strong>of</strong> re-positioning a 

subject‘s body in the calibration posture between acquisitions, can be a 

potential source <strong>of</strong> intra- and inter-examiner inaccuracy. A very similar 

problem, however, exists also for the <strong>protocols</strong> <strong>based</strong> on the identification <strong>of</strong> 

anatomical landmarks [6, 9, 19], and was named ―anatomical landmarks 

mislocation‖. The anatomical landmarks mislocation mostly affects the <strong>of</strong>fset 

<strong>of</strong> segment axial rotations [11], and even though final conclusions require adhoc 

tests (which are underway), this might also be true for Outwalk. 

As element supporting Outwalk validity, the inter-rater reliability for h and k 

was found to be very high, with values for hRMS 

and k RMS smaller than 

those reported in [11] for the same measurements obtained through an 

optoelectronic systems and the CAST <strong>protocols</strong>. Hypothesis H1 was thus 

196

confirmed. Results for SEM se and SEM nse also indicate that the ‗typical error‘ in 

the <strong>of</strong>fset <strong>of</strong> hip and knee flexion angles during gait is almost negligible, and 

thus it is not expected to substantially depend on the rater leading the 

measurements. This conclusion is further supported by the close values <strong>of</strong> 

SEM se and SEM nse (no difference for h, 0.1° for k), which show that the 

systematic error introduced by T1 and T2 is very limited. The measurement setup, 

i.e. static measures with subjects lying on a mat, with the legs sustained by 

a foam cylinder, with hips and knees flexed, appears to have positively 

influenced the reliability <strong>of</strong> h and k. In this posture the anatomical landmarks 

can also be easily palpated and the mat can be used as a base for the goniometer 

in measuring h, which in fact resulted as the most reliable angle. 

In conclusion, the results obtained support the commencement <strong>of</strong> a clinical trial 

<strong>of</strong> Outwalk with children with CP. 

This further step will also take advantage <strong>of</strong> the possibility to apply Outwalk 

with any optoelectronic system as measurement device, and not only with 

IMMS. It should be noticed in fact that all available optoelectronic systems 

satisfy the hardware requirements described in section 2.1.2, if we 1) consider 

the SU as a cluster <strong>of</strong> at least three not aligned markers, and 2) we define a 

local CS for the cluster with a visible reference to its markers‘ position. The 

synchronous application <strong>of</strong> Outwalk and a reference clinical protocol (e.g. 

CAST [2]) with the optoelectronic system as only measurement device will 

allow, for instance, to compare the effect on joint kinematics <strong>of</strong> the different 

definitions <strong>of</strong> the anatomical/functional CSs among the two <strong>protocols</strong>. Based on 

the results on clinical populations, it could be even decided to use Outwalk as 

first screening gait protocol in combination with optoelectronic systems, due to 

its ease <strong>of</strong> use. 

197


[1] Andriacchi TP, Alexander EJ (2000) Studies <strong>of</strong> human loco<strong>motion</strong>: past, 

present and future. J Biomech 33(10): 1217-1224. 

[2] Benedetti MG, Catani F, Leardini A, et al (1998) Data management in gait 

<strong>analysis</strong> for clinical applications. Clin Biomech 13(3): 204-215. 

[3] Best R, Begg R (2006) Overview <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> and gait feature. 

Computational Intelligence for Movement Sciences. IGP, Begg R , 

Palaniswami M. Hershey, Pennsylviania, USA 

[4] Biomch-L (July-August 2007) Discussion about: "Functional methods for 

the ankle complex". 

[5] Bobath B, Bobath K (1975) Motor <strong>development</strong> in the different types <strong>of</strong> 

cerebral palsy. William Heinemann Medical Books, London 

[6] Cappozzo A, Catani F, Della Croce U, et al (1995) Position and orientation 

in space <strong>of</strong> bones during movement: anatomical frame definition and 

determination. Clin Biomech 10(4): 171-178. 

[7] Chang FM, Seidl AJ, Muthusamy K, et al (2006) Effectiveness <strong>of</strong> 

instrumented gait <strong>analysis</strong> in children with cerebral palsy--comparison <strong>of</strong> 

outcomes. J Pediatr Orthop 26(5): 612-616. 

[8] Cutti AG, Giovanardi A, Rocchi L, et al (2008) Ambulatory measurement 

<strong>of</strong> shoulder and elbow kinematics through inertial and magnetic sensors. Med 

Biol Eng Comput 46(2): 169-178. 

[9] Davis RB III, Ounpuu S, Tyburski D, et al (1991) A gait data collection and 

reduction technique. Hum Mov Sci 10: 575-587. 

[10] De Vet HCW, Terwee CB, Knol DL, et al (2006) When to use agreement 

versus reliability measures. Journal <strong>of</strong> Clinical Epidemiology 59: 1033-1039. 

[11] Della Croce U, Cappozzo A, Kerrigan DC (1999) Pelvis and lower limb 

198

anatomical landmark calibration precision and its propagation to bone geometry 

and joint angles. Med Biol Eng Comput 37: 155-161. 

[12] Ferrari A, Cutti AG, Gar<strong>of</strong>alo P, et al (2009) First in-vivo assessment <strong>of</strong> 

‗outwalk‘ – a novel protocol for clinical gait <strong>analysis</strong> <strong>based</strong> on inertial & 

magnetic sensors. Med Biol Eng Comput; accepted 

[13] Ferrari Ad (1993) The use <strong>of</strong> epidemiology in disabilities: criteria for 

classification. Bottos M, Scrutton D, Ferrari A, Neville BGR (Eds.) The 

restored infant, 16-20 Firenze: Fisioray 

[14] Ferrari Ad, Alboresi S, Muzzini S, et al (2008) The term diplegia should 

be enhanced. Part I: a new rehabilitation oriented classification <strong>of</strong> cerebral 

palsy. Eur J Phys Rehabil Med 44(2): 195-201. 

[15] Kaufman KR, Levine JA, Brey RH, et al (2007) Gait and balance <strong>of</strong> 

transfemoral amputees using passive mechanical and microprocessor-controlled 

prosthetic knees. Gait Post 26: 489-493. 

[16] Kavanagh JJ, Menz HB (2008) Accelerometry: a technique for quantifying 

movement patterns during walking. Gait Post 28: 1-15. 

[17] Kontaxis A, Cutti AG, Johnson G, et al (2009) A framework for the 

definition <strong>of</strong> standardized <strong>protocols</strong> for measuring upper-extremity kinematics. 

Clin Biomech 24(3): 246-253. 

[18] O‘Donovan KJ, Kamnik R, O‘Keeffe DT, et al (2007) An inertial and 

magnetic sensor <strong>based</strong> technique for joint angle measurement. J Biomech 40: 

2604-2611. 

[19] Picerno P, Cereatti A, Cappozzo A (2008) Joint kinematics estimate using 

wearable inertial and magnetic sensing modules. Gait Post 28(4): 588-595. 

[20] Ramsey DK, Wretenberg PF (1999) Biomechanics <strong>of</strong> the knee: 

methodological considerations in the in vivo kinematic <strong>analysis</strong> <strong>of</strong> the 

tibi<strong>of</strong>emoral and patell<strong>of</strong>emoral joint. J Biomech 14: 595-611. 

[21] Roetenberg D (2006) Inertial and Magnetic Sensing <strong>of</strong> Human Motion. 

199

PhD Thesis Twente University 

[22] Sabatini AM (2006) Inertial sensing in biomechanics: a survey <strong>of</strong> 

computational techniques bridging <strong>motion</strong> <strong>analysis</strong> and personal navigation. 

Computational Intelligence for Movement Sciences, PM Begg R. IGP 

[23] Schache AG, Baker R, Lamoreux LW (2006) Defining the knee joint 

flexion–extension axis for purposes <strong>of</strong> quantitative gait <strong>analysis</strong>: An evaluation 

<strong>of</strong> methods. Gait Post 24: 100-109. 

[24] Schmalz T, Blumentritt S, Jarasch R (2002) Energy expenditure and 

biomechanical characteristics <strong>of</strong> lower limb amputee gait: the influence <strong>of</strong> 

prosthetic alignment and different prosthetic components. Gait Post 16: 255- 

263. 

[25] Schwartz MH, Rozumalski A (2005) A new method for estimating joint 

parameters from <strong>motion</strong> data. J Biomech 38: 107-116. 

[26] Sciavicco L, Siciliano B (2000) Modelling and control <strong>of</strong> robot 

manipulators. Springer-Verlag 

[27] Stagni R, Fantozzi S, Cappello A (2006) Propagation <strong>of</strong> anatomical 

landmark misplacement to knee kinematics: performance <strong>of</strong> single and double 

calibration. Gait Post 24(2): 137-141. 

[28] Sutherland DH, Davids JR (1993) Common gait abnormalities <strong>of</strong> the knee 

in cerebral palsy. Clin Orthop Relat Res 288: 139-147. 

[29] Weir JP (2005) Quantifying test-retest reliability using the intraclass 

correlation coefficient and the SEM. J Strength Cond Res 19(1): 231-240. 

[30] Winters F, Gage JR, Hicks R (1987) Gait patterns in spastic hemiplegia in 

children and young adults. JBJS 69-A(3): 437-441. 

[31] Woltring HJ (1990) Data processing and error <strong>analysis</strong>. Cappozzo A, 

Berme N (Eds), Biomechanics <strong>of</strong> human movement 

[32] Wu G, Siegler S, Allard P, et al (2002) ISB recommendation on definitions 

200

<strong>of</strong> joint coordinate system <strong>of</strong> various joints for the reporting <strong>of</strong> human joint 

<strong>motion</strong>. Part I. Ankle, hip and spine. J Biomech 35(4): 543–548. 

201


The following sections describe the validation <strong>of</strong> Outwalk protocol on 

transtibial amputees and the application <strong>of</strong> a novel method (KiC) for the 

description <strong>of</strong> the 3D kinematics <strong>of</strong> transfemoral amputees. 

202

4.2.1 Validation <strong>of</strong> Outwalk protocol on below-knee amputees 

3D GAIT KINEMATIC OF TRANSTIBIAL AMPUTEES 

WALKING IN EVERY-DAY-LIFE ENVIRONMENTS: 

RELIABILITY STUDY OF A PROTOCOL BASED ON 

INERTIAL & MAGNETIC SENSORS 

Cutti AG, Raggi M, Gar<strong>of</strong>alo P, Bottoni G, Amoresano A 

Accepted as poster at ISPO 2010, 10-15 May 2010, Leipzig (Germany) 

Summary 

A protocol named Outwalk has been recently proposed for the 3D gait <strong>analysis</strong> 

in real-life environments <strong>of</strong> transtibial amputees. This study addresses 

Outwalk‘s inter-rater reliability by involving 10 amputees and 2 rater. Results 

support the applicability <strong>of</strong> Outwalk in the clinical routine. 


The instrumental 3D gait <strong>analysis</strong> <strong>of</strong> amputees is currently limited to few 

prosthetic centres in which expensive movement <strong>analysis</strong> laboratories are 

available. Moreover, in the lab the gait <strong>of</strong> a patient can be conditioned by the 

stress imposed by the operators and by the artificial surrounding environment. 

Inertial and Magnetic Measurement Systems (IMMSs) might allow to 

overcome these limitations, being low-cost and portable. In addition, since the 

3D orientation <strong>of</strong> their Sensing Units (SU) is known in a global earth-<strong>based</strong> 

coordinate system, which is ubiquitous, long measurements can be possible 

―out-<strong>of</strong>-the-lab‖, in real-life environment, e.g. where the gait-training is carried 

out. For this purpose, we proposed a protocol named ‗Outwalk‘ to measure the 

3D kinematics <strong>of</strong> gait <strong>based</strong> on the IMMS by Xsens Technologies (NL). The 

aim <strong>of</strong> the present work was to test the inter-rater reliability <strong>of</strong> the protocol on 

Transtibial Amputees (TA). 

Methods 

203

To measure the pelvis-trunk, hips, knees, and ankles 3D kinematics, Outwalk 

requires to 

1) position 8 SUs on trunk and lower-limb segments; 

2) flex-extend each knee to estimate its mean flexion-extension rotation axis; 

3) measure the SUs‘ orientation with the subject in the upright anatomical 

posture (See Paragraph 4.1.2 <strong>of</strong> this thesis). 

Ten TA (45±10 year-old, K2-K3 level) participated in the experiment after 

signing an informed consent, together with 2 operators (O1, O2). O1 and O2 

independently applied Outwalk on each subject and acquired the amputee‘s gait 

kinematics while walking at self-selected speed in the park <strong>of</strong> our Centre along 

a 30m straight path. Acquisitions by O1 and O2 were 10 min apart. Gait cycles 

were segmented using the algorithm described in [1] To quantify the interoperator 

reliability we computed, among others, the Standard Error <strong>of</strong> 

Measurement (SEM) <strong>of</strong> the 36 parameters described in [2] , <strong>based</strong> on an 

ANOVA with repeated measures, as recommended in [3,4] . 

Results 

For the interest <strong>of</strong> brevity, Table 1 reports SEM values for the 14 most 

significant parameters <strong>of</strong> the 36 examined, both for the sound and prosthetic 

side. The SEMs reported both consider random and systematic effects. The 

names used for the parameters are those reported in [2] , to which the reader is 

referred for a detailed description. Here suffices to say that: 1) H, K, A refer to 

hip, knee and ankle; 2) parameters ending with 6 and 7 refer to the sagittal and 

frontal plane range <strong>of</strong> <strong>motion</strong> (ROM); 3) ending with 2 refer to the maximum 

flexion/plantaflexion at loading response; ending with 3 refer to the maximum 

extension/dorsiflexion in stance phase; ending with 5 refer to the maximum 

flexion/dorsiflexion in swing. 

H3 H5 H6 H7 K2 K3 K5 K6 K7 A2 A3 A5 A6 

Sound 2.8 2.7 0.7 1.3 1.9 2.0 2.0 1.9 2.9 1.7 1.8 2.5 1.4 

Affected 2.1 2.8 1.7 1.0 1.6 0.7 1.9 1.4 3.4 0.9 0.8 0.9 0.5 

Table 1 SEM values for important features <strong>of</strong> the kinematic patterns <strong>of</strong> hip, knee and ankle, as 

defined in [3] 

204

Conclusion 

Results appear consistent with reports on other populations [3,5] . In particular, 

the sagittal ROMs (H-K-A6) have a SEM

4.2.2 Evaluation <strong>of</strong> above-knee amputees kinematics during 

gait using inertial sensors 

LOWER LIMB 3D JOINT KINEMATICS MEASUREMENT 

ON ABOVE-KNEE AMPUTEES DURING GAIT USING 

INERTIAL AND MAGNETIC MEASUREMENT SYSTEMS 

To be submitted 


Lower limb 3D joint kinematics <strong>of</strong> above-knee amputees is typically measured 

by means <strong>of</strong> optoelectronic systems. Although very accurate and established in 

literature as the golden standard, these systems cannot be easily adopted as 

ambulatory measurement tools, in free-living conditions. In order to support 

orthopaedic technicians and clinicians in the evaluation <strong>of</strong> prosthetic devices 

and in monitoring the improvements <strong>of</strong> the subject during the rehabilitation 

process, portable <strong>motion</strong> capture systems using Inertial and Magnetic 

Measurement Systems (IMMS) may represent a valid option. 

Recently a protocol named Outwalk was developed to easily measure on 

amputees the thorax-pelvis and lower-limb 3D kinematics during gait (Chapter 

4) by means <strong>of</strong> IMMS, such as MTx (Xsens Technologies B.V., The 

Netherlands). Although the protocol was developed taking into account the 

target population <strong>of</strong> lower limb amputees, the validation was only carried out 

on healthy subjects. 

The presence <strong>of</strong> ferromagnetic materials inside <strong>of</strong> the lower limb prostheses 

may limit the use <strong>of</strong> this protocol through the IMMS for representing the proper 

prosthetic limb kinematics and the evaluation <strong>of</strong> the unimpaired limb, which is 

also important for supporting the rehabilitation treatment and the design <strong>of</strong> 

prostheses. The measurement <strong>of</strong> the earth magnetic field by IMMS can be 

affected by ferromagnetic materials inside <strong>of</strong> the prosthetic devices. This has 

direct consequences on the heading estimation. Moreover, non-homogeneity <strong>of</strong> 

the earth magnetic field may occur depending on the environment in which the 

clinical trials have to be carried out, such as laboratories with common 

instrumentation, corridors with common furniture, walking paths outside <strong>of</strong> the 

laboratories. 

206

In this work, the limitations in the 3D joint kinematics measurement on aboveknee 

amputees by means <strong>of</strong> IMMS were examined and discussed. 

The objectives were (1) to evaluate the magnetic field distortions occurring 

during walking when an above-knee amputee is fitted with an electronic knee 

(C-Leg, Ottobock Healthcare, Germany), (2) to examine the consequences <strong>of</strong> 

the non-homogeneity <strong>of</strong> the earth magnetic field on the 3D joint kinematics <strong>of</strong> 

the prosthetic and unimpaired limbs, and finally (3) to test the KiC algorithm, a 

new method for improving the accuracy and correctly representing the 3D joint 

kinematics when the earth magnetic field is not homogeneous. 

2. METHODS 

An above-knee amputee fitted with a C-Leg electronic knee prosthesis 

participated in the experiment. A Vicon optoelectronic <strong>motion</strong> capture system 

was adopted as the reference system and the MTx Xsens system as IMMS. 

Hence, all the measurements inside <strong>of</strong> the laboratory were performed running 

Xsens and Vicon simultaneously like described in [6]. 

In Figures 1 and 2 a schematization <strong>of</strong> the setup for the unimpaired limb and 

the prosthetic limb for the calculation <strong>of</strong> knee joint kinematics, is shown. In the 

figures, the distances between the centroids <strong>of</strong> the cluster <strong>of</strong> markers and the 

knee joint are indicated. The vector indicating the position <strong>of</strong> the knee joint 

expressed in the local reference frame <strong>of</strong> the sensing unit is, for each sensing 

unit, the input provided to the KiC algorithm for the estimation <strong>of</strong> the relative 

heading. 

Position <strong>of</strong> the centroids were calculated through the marker positions, the knee 

joint was estimated through the calculation <strong>of</strong> the external anatomical 

landmarks through Vicon. In the procedure which follows, the centroid <strong>of</strong> the 

cluster <strong>of</strong> markers and the origin <strong>of</strong> the sensing unit will be assumed coincident. 

207

Figure 1 – Schematization <strong>of</strong> the unimpaired limb for the calculation <strong>of</strong> the knee 3D joint 

Kinematics 

208

Figure 2 – Scheamtization <strong>of</strong> the prosthetic limb for the calculation <strong>of</strong> the knee 3D joint 

Kinematics 

A mapping <strong>of</strong> the magnetic field inside <strong>of</strong> the laboratory was performed. 

209

In Figure 3, a schematization <strong>of</strong> the basic approach adopted for achieving the 

objectives <strong>of</strong> the <strong>analysis</strong> is showed. 

Figure 3– Schematization <strong>of</strong> the approach followed for the segment kinematics comparison. 

During each dynamic trial, a ―virtual xsens cluster‖ is created. The orientation 

<strong>of</strong> the ―virtual Xsens cluster‖ is the orientation <strong>of</strong> the rigid body which includes 

the Vicon cluster <strong>of</strong> markers and the Xsens sensing unit, constantly aligned to 

the Xsens embedded frame and described by the technical frame <strong>of</strong> the cluster 

<strong>of</strong> markers. In other words, the movement <strong>of</strong> the virtual xsens cluster 

corresponds to the movement <strong>of</strong> the Xsens unit but described using Vicon 

orientation. 

For the sake <strong>of</strong> clearness, hereinafter the procedure will be described by 

rotation matrix instead <strong>of</strong> quaternion. For each frame t <strong>of</strong> the dynamic trial, the 

rotation matrix representing the orientation <strong>of</strong> the virtual xsens cluster over the 

thigh segment is defined as: 

= (1) 

represents the orientation <strong>of</strong> the cluster <strong>of</strong> markers over the thigh 

(TH) segment in the Vicon global reference frame. 

is the hand-eye calibration matrix [8] which is the constant relation 

between the Xsens local coordinate system and the local coordinate system 

created from the Vicon cluster <strong>of</strong> markers. During the hand-eye calibration 

procedure, 3D orientation <strong>of</strong> the sensing unit in the Xsens global reference 

frame was calculated using the XKF3 algorithm provided with the Xbus kit 

210

system. 

The virtual Xsens cluster and the Xsens orientations are compared in terms <strong>of</strong> 

relative <strong>motion</strong>, as follows. 

First, from (1) the orientation <strong>of</strong> the virtual cluster <strong>of</strong> the thigh segment 

is extracted at the instant 

: 

with respect to the Vicon global reference system 

Then the relative orientation <strong>of</strong> the virtual cluster with respect to the orientation 

at the instant , is computed for each instant <strong>of</strong> the dynamic task: 

= * (2) 

The same procedure is adopted for the Xsens unit, that is in the same dynamic 

task, from (1) the orientation <strong>of</strong> the Xsens unit at the instant with respect 

to the Xsens global reference system is extracted: 

Now the relative orientation <strong>of</strong> the Xsens unit with respect to the orientation at 

the instant , for each instant <strong>of</strong> the dynamic task can be calculated: 

= * (3) 

The final comparison is therefore performed between (2) and (3), converting 

orientation data into quaternion in order to avoid singularities and extracting the 

parameters RMSE, the coefficient <strong>of</strong> correlation r and the parameters <strong>of</strong> the 

regression line, m and q, following the same technique and interpretation 

criteria described in [7]. 

The same procedure is then adopted for the shank (SH) and foot (FT) segments. 

211

2.1 Characterization <strong>of</strong> the magnetic field with C-Leg prosthesis 

Objective 1 was accomplished by evaluating the MTx accuracy in the 

representation <strong>of</strong> segment kinematics for 5 walking cycles inside <strong>of</strong> the 

laboratory, using the optoelectronic system as a reference. 

For each body segment the variation in the magnetic field, measured by the 

magnetometers included in each MTx unit and the segment kinematics obtained 

from Vicon and Xsens were investigated. The magnetic field was represented 

as the amplitude <strong>of</strong> the magnetic field vector normalized with respect to the 

earth magnetic field. 

Segment kinematics was represented and compared adopting the above 

described method. 

2.2. Consequences on joint kinematics for the prosthetic limb (Vicon+CAST Vs 

Xsens+CAST) 

The comparison <strong>of</strong> the 3D joint kinematics was performed having the two 

systems sharing the same protocol <strong>based</strong> on CAST [9] . 

In Figure 4 and 5, a schematization <strong>of</strong> the comparison between Vicon and 

Xsens data when sharing the CAST protocol is shown. 

Figure 4 – Schematization <strong>of</strong> the determination <strong>of</strong> the calibration matrix needed for applying 

CAST on Xsens data 

During a static calibration task (Figure 4), for each body segment, CAST 

technique is applied on the Vicon cluster <strong>of</strong> markers technical frame obtaining 

the anatomical frame expressed in the Vicon global reference frame. 

The hand-eye calibration matrix is applied to the vicon cluster technical frame 

obtaining the virtual Xsens cluster previously described. 

Then, the relative orientation between the anatomical frame and the virtual 

Xsens cluster frame is calculated. This matrix will be used to perform the 

sensors-to-segment calibration allowing to describe the anatomical frame 

212

through the Xsens embedded frame. 

In Figure 5 the procedure adopted for the application <strong>of</strong> the sensor-to-segment 

calibration and the final comparison is showed, in the case <strong>of</strong> knee joint 

kinematics. 

Figure 5 – Schematization <strong>of</strong> the procedure followed for the joint kinematics comparison between 

Xsens and Vicon 

During each dynamic trial, CAST technique is applied to the thigh, shank 

segment cluster <strong>of</strong> markers, obtaining the corresponding anatomical frames. 

The relative orientation between the virtual Xsens cluster and the anatomical 

frame was previously obtained for the thigh and the shank segment. Now, this 

relative orientation is applied to the Xsens orientation, obtaining the ―Xsens 

anatomical frame‖ which is the orientation <strong>of</strong> the Xsens unit aligned to the 

CAST anatomical frame. 

During dynamic trials, 3D orientation <strong>of</strong> the sensing unit in the Xsens global 

reference frame was calculated using the XKF3 algorithm provided with the 

Xbus kit system. 

The relative orientation between the thigh and the shank is separately calculated 

for a) the thigh and shank anatomical frames obtained through Vicon and b) the 

thigh and the shank anatomical frames obtained through Xsens. 

The resulting Euler angles from a) and b) are finally compared. 

The same procedure is applied for the shank and foot segments in order to 

compare the ankle joint kinematics obtained through Vicon and Xsens systems. 

2.3 Test <strong>of</strong> a new method for improving the accuracy and correctly 

representing the 3D joint kinematics 

Objective 3 was accomplished by testing a new algorithm, named KiC 

213

(Kinematic Coupling) developed by Xsens Technologies B.V.. In order to 

evaluate if KiC could improve the accuracy <strong>of</strong> MTx and correctly represent the 

kinematics <strong>of</strong> the prosthetic limb, when distortions in the magnetic field 

affected the 3D orientation estimation, the same procedure described in 2.2 was 

applied using KiC instead <strong>of</strong> XKF3. This operation was performed both for the 

walking trials inside <strong>of</strong> the laboratory and a long walking trial out <strong>of</strong> the 

laboratory. Note that being the sensor-to-segment calibration a constant relation 

between the sensing units positioned over the body segments and the 

anatomical frames, it can also be applied out <strong>of</strong> the laboratory. 

Both for the inside and outside walking trials, the first reference for 

understanding which result is reasonable, is due to the inner nature <strong>of</strong> the knee 

and foot prostheses: the knee joint kinematics is expected to be the result <strong>of</strong> a 

single rotation as for a hinge joint, as well as for the ankle. 

2.3.1 Description <strong>of</strong> KiC 

For a description <strong>of</strong> KiC and how it can be interfaced with CAST protocol, 

refer to Chapter 6 <strong>of</strong> this thesis. 

3.RESULTS 

Inside walking trials 

Segment kinematics comparison among the 5 walking trials between Xsens and 

Vicon showed mean RMSE values for thigh and shank segments <strong>of</strong> the 

prosthetic limb <strong>of</strong> respectively 2.4° ± 1.2° and 2.3° ± 1.8°. It is important to 

notice that these results are related to the individual segments: the error in the 

relative orientation kinematics can be amplified by the combination <strong>of</strong> the 

errors in the two segment orientations. Mean values <strong>of</strong> coefficient <strong>of</strong> correlation 

[7] 

are close to 1. Hence, the high values <strong>of</strong> RMSE are explained by the 

presence <strong>of</strong> an <strong>of</strong>fset between the orientations provided by the two systems. 

During the walking trials inside <strong>of</strong> the laboratories, the magnetic field 

distortions are noticeable for the sensing units over the shank and foot 

segments, as shown in Figure 6 and 7. For the shank segment, the norm <strong>of</strong> the 

magnetic field vector (represented by the black curve and being 1 the default 

value) is variable around the interval -1.5 and +1.5. For the foot the same vector 

is variable around -1.5 and 2. It must be pointed out that the final output <strong>of</strong> 

XKF3 contains the combination <strong>of</strong> the effects produced by the distortions 

measured by both the proximal and distal segment. For what concerns the knee 

214

kinematics, the magnetic distortions occurring at the shank segment are the 

main source <strong>of</strong> error. 

Figure 6 – Accelerations, angular velocities, magnetif field for the shank segmento <strong>of</strong> the 

prosthetic limb. The black curve corresponds to the norm <strong>of</strong> the magnetic field vector. The ideal 

value is 1. 

Figure 7 - Accelerations, angular velocities, magnetif field for the foot segmento <strong>of</strong> the prosthetic 

limb. The black curve corresponds to the norm <strong>of</strong> the magnetic field vector. The ideal value is 1. 

A typical example <strong>of</strong> comparison between the knee and ankle joint kinematics 

215

for the prosthetic limb obtained through Vicon and the one obtained through 

KiC, sharing CAST, i.e. after the application <strong>of</strong> the sensor-to-segment 

calibration, is shown in figures 8 and 9. 

Figure 8 –Ankle joint kinematics comparison between Vicon (blue curves) and Xsens KiC (orange 

curves), for the prosthetic limb 

Figure 9 – Knee joint kinematics comparison between Vicon (blue curves) and Xsens KiC (orange 


216

Figures 10 and 11 show the comparison between the knee and ankle joint 

kinematics for the prosthetic limb obtained through XKF3 and the one obtained 

through the KiC algorithm, sharing CAST, i.e. after the application <strong>of</strong> the 

sensor-to-segment calibration. The overall pattern recorded through Xsens is 

shown, therefore including the first minute <strong>of</strong> settling time for XKF3. Despite 

<strong>of</strong> the initialization time, the output <strong>of</strong> XKF3, which is consequence <strong>of</strong> using 

the magnetic sensors, seems affected by the magnetic distortion due to the 

prosthesis. 

Figure 10 – Knee joint kinematics comparison between XKF3 (blue curves) and KiC (red curves), 

for the prosthetic limb 

217

Figure 11 – Ankle joint kinematics comparison between XKF3 (blue curves) and KiC (red 


Outside walking trials 

Out <strong>of</strong> the laboratory, knee kinematics <strong>of</strong> the unimpaired limb, resulting from 

the application <strong>of</strong> CAST to Xsens presented patterns consistent with results 

from literature [2] . For the prosthetic limb, knee kinematics presented large 

confidence bands for ab-adduction and internal-external rotation angles, 

respectively from -12.5° to 12.5° and from – 15° to + 15° . 

218

Figure 12 – Knee joint angles comparison between Xsens+KiC+CAST and Vicon+CAST, for the 

prosthetic side 

When applying the kinematic coupling method, the bands for ab-adduction and 

internal-external rotation were restricted to [ -2° , 0.7° ] and [ -2.3° , 1.7° ] . 

In Figure 12, differences in knee joint angles when comparing Xsens and Xsens 

after applying KiC, sharing the CAST technique, are presented during the 

outside walking trials. 

4. CONCLUSIONS 

There were several advantages in using a protocol <strong>based</strong> on CAST. First <strong>of</strong> all, 

CAST is a technique commonly adopted among biomechanists and it allows to 

estimate accurate 3D joint kinematics. Secondly, CAST technique provides 

information about the position <strong>of</strong> landmarks which was adopted for the test <strong>of</strong> 

the KiC algorithm as third objective <strong>of</strong> this work. Finally, the creation <strong>of</strong> the 

anatomical frames using CAST during tha static calibration is independent 

from the magnetic field. Therefore, no errors due to the magnetic distortions are 

propagated throughout the data processing during the sensor-to-segment 

calibration. 

Although the estimation <strong>of</strong> the virtual cluster orientation depends on the handeye 

calibration in which the Xsens system is involved, a mapping <strong>of</strong> the 

219

laboratory excluded magnetic distortions occurring during the procedure. 

Despite <strong>of</strong> this, errors in the estimation <strong>of</strong> the points and so in the creation <strong>of</strong> 

the anatomical frames can occur. They depend on the s<strong>of</strong>t tissue artifact, 

palpation <strong>of</strong> the landmarks by the operator. 

Results for the prosthetic limb during inside walking confirmed the inaccuracy 

in the segment kinematics estimation, due to magnetic distortions. 

Although knee flexion-extension angle seems to be well represented, it is worth 

to notice that the prosthetic knee joint is designed as an hinge joint, i.e. abadduction 

and internal-external rotation angles must be theoretically null. 

XKF3 did not provide a valid output both for knee and ankle kinematics. 

The shank and foot segments orientation seem to be the most affected by the 

magnetic distortions. 

KiC was proved to be effective in improving the accuracy <strong>of</strong> knee and ankle 

kinematics representation during gait. The small errors in the angle components 

out <strong>of</strong> the sagittal plane for knee and ankle joints are within the instrumental 

error <strong>of</strong> Xsens (dynamic accuracy <strong>of</strong> 2 degrees <strong>of</strong> RMS). 

Differences between Vicon and Xsens for the inside walking trials cannot be 

explained by the sensor-to-segment calibration procedure, being it in common 

to both the systems. The differences are probably due to filtering effects 

occurring adopting automatic procedures like Woltring filtering [10] to Vicon 

data. 

220



assessment <strong>of</strong> gait temporal symmetry in above-knee amputees: The ‗SEAG‘ 

protocol. Gait Posture. 2008;28:S31-S32. 

2. Benedetti MG, Catani F, Leardini A, Pignotti E, Giannini S. Data 

management in gait <strong>analysis</strong> for clinical applications. Clinical Biomechanics. 

1998;13(3):204-215. 

3. Weir JP. Quantifying test-retest reliability using the intraclass correlation 

coefficient and the SEM. J Strength Cond Res. 2005;19(1):231-240. 

4. Jennifer L. McGinley, Richard Baker, Rory Wolfe, Meg E. Morris. The 

reliability <strong>of</strong> three-dimensional kinematic gait measurements: A systematic 

review. Gait Posture. 2009;29(3):360-369. 

5. Fortin C, Nadeau S, Labelle H. Inter-trial and test–retest reliability <strong>of</strong> 

kinematic and kinetic gait parameters among subjects with adolescent 

idiopathic scoliosis. European Spine Journal. 2008;17(2):204-216. 

6. Ferrari A, Cutti A, Gar<strong>of</strong>alo P, et al. First in vivo assessment <strong>of</strong> "Outwalk": a 

novel protocol for clinical gait <strong>analysis</strong> <strong>based</strong> on inertial and magnetic sensors. 

Med Biol Eng Comput. 2009. Available at: 

http://www.ncbi.nlm.nih.gov/pubmed/19911215 [Accessed December 4, 2009]. 




8. Ferrari A, Cutti A, Gar<strong>of</strong>alo P, et al. First in vivo assessment <strong>of</strong> "Outwalk": a 

novel protocol for clinical gait <strong>analysis</strong> <strong>based</strong> on inertial and magnetic sensors. 

Med Biol Eng Comput. 2009. Available at: 

http://www.ncbi.nlm.nih.gov/pubmed/19911215 [Accessed December 4, 2009]. 



221


10. Woltring H. Data processing end error <strong>analysis</strong>. Biomechanics <strong>of</strong> human 

movements. Applications in rehabilitation, sports and ergonomics. Chapter 10. 

222


SOFTWARE FOR THE PROTOCOLS BASED ON 


223

4.3.1 Design <strong>of</strong> Outwalk Manager and main features 

DEVELOPMENT OF A CLINICAL SOFTWARE TO 

MEASURE THE 3D GAIT KINEMATICS IN EVERY-DAY- 

LIFE ENVIRONMENTS THROUGH THE OUTWALK 

PROTOCOL 

Gar<strong>of</strong>alo P, Raggi M, Ferrari A, Cutti AG, Davalli A 

Gait & Posture, vol. 30, suppl. 1, p. 30 (October 2009) 


A protocol named ‗Outwalk‘ was recently proposed to measure the thorax, 

pelvis and lower-limb 3D kinematics during gait in real-life environments, by 

means <strong>of</strong> the Xsens IMMS (Inertial and Magnetic Measurement System) 

(Xsens Technologies, NL). Outwalk was validated against a clinical reference 

protocol (CAST) with positive results. 

The protocol requires the execution <strong>of</strong> 2 main steps: 

a) static calibration, for the estimation <strong>of</strong> the sensor-to-segment calibration 

matrices 

b) functional calibration, for the estimation <strong>of</strong> the functional knee flexion 

extension axis 

The subject is then ready for walking. Finally a standard clinical report is 

created. To make Outwalk really effective in the clinical practice, it has to be 

made available to the end-user through an easy-to-use s<strong>of</strong>tware. The aim <strong>of</strong> this 

work was to develop such a s<strong>of</strong>tware, named ‗Outwalk Manager‘. 

Methods 

Outwalk Manager was developed using Matlab GUI Builder (The Mathworks, 

USA) and then compiled as stand-alone application. As additional tool, to 

224

establish the Bluetooth communication with the Xsens system and to compute 

the real-time 3D orientation <strong>of</strong> its sensing units (SUs) through a Kalman filter 

(XKF3) (Xsens Manual), the CMT COM Object provided in the Xsens SDK 

was used. The s<strong>of</strong>tware architecture comprises 9 components: 

1) association <strong>of</strong> the SUs to the body segments; 

2) patient database; 

3) measurement settings, to set the sample frequency (from 60 to 120 Hz) and 

XKF3 scenario; 

4) measurement activation and recording <strong>of</strong> the raw accelerometric, gyroscopic 

and magnetometric data and the 3D SUs orientation data; 

5) Kalman filter settling and filter state saving; 

6) ―sensor-to-segment calibration‖ with the definition <strong>of</strong> the 

anatomical/functional coordinate systems; 

7) real-time visualisation <strong>of</strong> the‖ joint-angles vs time‖ or ―joint-angle vs jointangle‖ 

plots [2], with the opportunity to compare real-time data with data 

previously recorded; 

8) gait cycles segmentations through the SEAG algorithm, which detects the 

gait events <strong>based</strong> on the shanks angular velocity; 

9) generation <strong>of</strong> the standard clinical report. 

Results 

Operatively, after establishing the connection with the system and assigning the 

SUs to the body segments, the user is asked to enter the patient‘s description. 

Then, the user can change the default measurement parameters (100Hz, 

―human large acceleration‖ scenario) and put the system into the measurement 

modality. After 2 minutes <strong>of</strong> Kalman filter settling, the user can proceed with 

the execution <strong>of</strong> the ―sensor-to-segment calibrations‖: static trial first, and then 

knee functional tasks. After that, the system is ready to record an indefinitely 

long trial, potentially comprising hundreds <strong>of</strong> gait cycles, presenting to the user 

one <strong>of</strong> the two real-time joint kinematics visualisation, joint-angles vs time and 

joint-angle vs joint-angle. After the execution <strong>of</strong> the trial, the SEAG algorithm 

is applied and the 3D kinematics is automatically presented subdivided in gaitcycles: 

from trial to a standard report in 3 ―clicks‖. A video-demo <strong>of</strong> the 

s<strong>of</strong>tware is available at http://www.inail-starter.org/downloads. 

225


The Outwalk Manager features were designed to guide and support the user 

during measurements in clinical settings. The execution <strong>of</strong> the Outwalk 

protocol takes advantage <strong>of</strong> the potentialities provided by the recent Xsens 

SDK: a) the Kalman filter state saving (and restoring) avoids to repeat the 

settling <strong>of</strong> the Kalman filter before each trial, speeding up considerably the 

acquisition time; b) the opportunity to select the filter scenario extends the use 

<strong>of</strong> the protocol for activities with different dynamics; c) IMMS raw data 

recording allows to have a complete overview for a posteriori evaluation <strong>of</strong> the 

quality <strong>of</strong> the measurement. 

4.3.2 Use <strong>of</strong> Outwalk Manager in clinical settings 

Outwalk protocol was proposes for measuring the thorax, pelvis and lower limb 

3D kinematics during gait real-life environments, by means <strong>of</strong> the Xsens 

IMMS (Xsens Technologies B.V. (The Netherlands). 

Simple steps (Figure 1) are required for the application <strong>of</strong> Outwalk protocol. 

Static and functional calibrations are necessary for the sensor-to-segment 

calibrations. 

Figure 1 – Steps <strong>of</strong> Outwalk protocol. After the static and functional calibration, walking trials can 

be performed and a standard clinical report is created. 

Outwalk Manager, previously described, allows to execute the steps above and 

provide a standard clinical report as final outcome. The overall procedure, 

including the measurement <strong>of</strong> several long walking trials takes not more than 

30 minutes. 

Once the subject has arrived, the MTx sensing units can be mounted on the 

226

subject body segments, according to Outwalk protocol. 

Some spotchecks before running the s<strong>of</strong>tware are provided: 

- Remember to check strap over the thighs segments in order to verify 

that it is well fasten 

- Check the sensing units over the shanks. The z axes <strong>of</strong> the sensing 

units (perpendicular to the casing) are used to create the anatomical 

coordinate systems <strong>of</strong> the shanks, thus they should lie in the frontal 

plane <strong>of</strong> the shanks. 

- Check the sensing unit over the pelvis. It must be horizontal with the 

cable pins pointing to the left side <strong>of</strong> the subject 

- Check the sensing unit over the thorax. It must be over the flat portion 

<strong>of</strong> the sternum 

Now the user can run INAIL Manager and select Outwalk Protocol 

(monolateral or bilateral version) among the ones available (Figure 2). 

Figure 2 – Starting screen <strong>of</strong> INAIL Manager where the protocol can be selected by the user 

The s<strong>of</strong>tware allows the user to scan the MTx units connected to the Xbus kit 

and provide the correct association between the last two digits <strong>of</strong> the MTx 

device ID numbers and the corresponding body segment (Figure 3). 

227

Figure 3 – Association <strong>of</strong> the sensing units to the corresponding body segment 

Now the user can insert the patient information through the button ―new 

patient‖ (Figure 4). 

228

Figure 4 – Creation <strong>of</strong> a new set <strong>of</strong> information for a new patient 

First, the Kalman filter algorithm adopted in MTx need to be settled. The 

duration <strong>of</strong> this step depends on the specific filter adopted (normally around 1 

minute). 

The user enables the "save filter state" function, presses ―Start Measuring‖, 

then ―Start plotting. 

During the settling period, typically, the subject is asked to stand still for 30 

seconds in an area in which the magnetic field is not distorted, the subject is 

asked to walk slowly for 15 – 20 meters. 

After the settling period, the user will press ―Stop Plotting‖ and the state <strong>of</strong> the 

filter will be stored. The GUI will automatically disable the ―Start Measuring‖. 

Hence the user can press again Start Measuring. Hereinafter, everytime the user 

press it, the filter state previously saved is restored. From now on during the 

session, the user has only to use the Start Plotting and Stop Plotting buttons. 

Now the GUI will guide the user through the next step, that is the sensor-tosegment 

calibration procedures. 

229

Functional calibration 

The estimation <strong>of</strong> the functional axis <strong>of</strong> flexion-extension (Mean Helical Axes) 

<strong>of</strong> the knee for the right and then for the left side can be obtained trying to 

passively or directly asking to the subject to perform knee flexion-extension 

movements. Typically it is convenient to leave the subject sitting on a chair. 5-6 

repetitions <strong>of</strong> flexion-extensions up to 60-70° are suggested. After each MHA 

computation the s<strong>of</strong>tware returns the deviation angle between the MHA 

estimated and the Instantaneous Helical Axes. A good estimation <strong>of</strong> MHA is 

typically reached when this angle is below 12. Note that this parameter does not 

give you an indication about which is the direction <strong>of</strong> the MHA. The latter can 

be checked simply through the kinematics obtained after calibration. 

Figure 5 – Main acquisition window <strong>of</strong> Outwalk Manager 

Static calibration 

A static acquisition can be performed from an upright (Figure 6) or supine 

position (Figure 7) when the first is not feasible (e.g. cerebral palsy children). 

230

Figure 6 – Static calibration posture 

Figure 7 – Supine position for static calibration 

In the upright posture, all the joints should keep at neutral orientation, thus back 

straight, looking forward, knee centre aligned to the ASIS and the line from the 

2nd metatarsal head to the calcaneus <strong>of</strong> the right foot parallel to the same line 

<strong>of</strong> the left foot. When adopting the supine position, the user must enable the 

―Static Horizontal‖ check, then be sure that all the joints are in the neutral 

position. When the knee and hip flexion-extension angles are different than the 

231

neutral values, the user can measure them by a goniometer and include these 

values in the interface (Figure 8). 

Figure 8 – Part <strong>of</strong> Outwalk Manager interface for the correction <strong>of</strong> the flexion-extension angles 

during supine static calibration 

Dynamic tasks 

The subject is asked to walk and by the Start Plotting and Stop Plotting buttons 

the user can visualize real-time 3D joint Kinematics during walking as a result 

<strong>of</strong> the calibrations already performed. 

During the execution <strong>of</strong> the dynamic trials, either the joint kinematics in time 

(Figure 9) can be visualized or the inter-joint coordination plot (Figure 10). 

Figure 9 - Real-time joint kinematics plot vs time visualization 

232

Figure 10 – Real-time inter-joint coordination plot visualization 

After the gait is acquired, the user goes to the ―SEAG Panel‖presses the 

―Calibration‖ button, select the trial that was just acquired, then presses the 

―Segmentation‖ button, and finally ―Report‖. A standard clinical report (Figure 

11) presenting the 3D joint kinematics for one or both limbs depending on the 

protocol selected is automatically created. 

233

Figure 11 – Standard Kinematics report as out come <strong>of</strong> Outwalk Manager 

234

4.3.3 Outwalk Manager Tutorial 

1. Welcome to INAIL Manager. This tutorial will guide you through the 

use <strong>of</strong> INAIL Manager and Outwalk Protocol in particular. 

2. First, press "Scan" to connect to your system and checking for the 

MTx units available. 

3. Now, select the corresponding body segment for each MTx unit. 

4. Press "Next" to go to the main acquisition interface 

5. Press "Change Folder" to select the main folder in which all patients 

data will be stored 

6. Press "New patient" to create a new patient in the database. 

7. If you want to select a patient previously acquired, press "Find other 

patient" button 

8. Before proceeding with the protocol steps, let's have a brief 

introduction about how to control the measurement through the 

interface buttons. 

9. In the lower left corner, the "Control Panel" allows you to control the 

connection between the system and the data visualization. 

10. "Start Measuring" changes the modality <strong>of</strong> the system into 

Measurement Mode, while with "Stop Measuring" the system enters 

the Configuration Mode. 

11. In the upper right corner, "Xbus Master Settings" panel allow you to 

control the acquisition parameters. These settings will be applied only 

if the "Start Measuring" button is pressed 

12. "Start plotting" and "Stop Plotting" control the data visualization. Joint 

angles are displayed in terms <strong>of</strong> the three angle components. 

235

13. Press 'n' to change the visualized joint. 

14. Press 'z' to zoom in the signals. Press 'z' again to restore the normal 

visualization. 

15. "Save filter state" can be pressed before or during the data 

visualization to store the state <strong>of</strong> the system. The state will be restored 

everytime "Start Measuring" button is pressed. 

16. "Rescan" button can be used to go back to the previous interface, for 

changing the body segments order or to start again the communication 

with the system when it was accidentally lost. 

17. If you just entered the acquisition interface you must perform the 

initialization <strong>of</strong> the system before proceeding. 

18. Press "Start Measuring". The system is now in Measurement Mode 

and all the settings you selected through the interface were applied to 

the system. 

19. Wait about 1 minute. During this period, the subject should stand still 

for 40 seconds and than move around for some seconds and stand still 

again. 

20. If you do not want to save the state <strong>of</strong> your system, just leave the 

system in Measurement Mode. 

21. The first step <strong>of</strong> Outwalk protocol is to perform the estimation <strong>of</strong> the 

flexion-extension functional axis <strong>of</strong> the knee. 

22. Press "Start Plotting" to visualize the 3D knee joint angle. Ask the 

subject to perform some consecutive pure knee flexion-extension, with 

some short breaks between the flexions and the extensions. 

23. Then close the visualization. The system will indicate the "dispersion 

angle" parameter. Check if your results is around 10 degree and then 

save your data. 

236

24. If you want, you can perform, visualize and save more trials and then 

in the end choose the best one. 

25. If you are acquiring both the limbs, follow exactly the same 

instructions as before, for the left side. 

26. The second step <strong>of</strong> Outwalk protocol is to perform a static calibration 

posture. 

27. Press "Start Plotting" to visualize real-time joint angles during the 

static calibration posture. 

28. Wait at least 8 seconds before pressing "Stop plotting" or the X in the 

upper corner. 

29. If you want to perform a supine static calibration trial, first select the 

corresponding button. 

30. Then enter the values <strong>of</strong> the knee, hip and ankle angles during the 

static trial. 

31. Following the same instructions as before, you can now visualize data 

during the static trial. 

32. You are now ready to measure every kind <strong>of</strong> lower limb activity. 

33. Just enter the name <strong>of</strong> the trial and press "Start Plotting" to get realtime 

3D joint angles. 

34. If you want to visualize your data using angle-angle plots, first click 

on "Enable coordination plots", and press "Start Plotting". 

35. In the angle-angle plot interface, the acquisition time is displayed in 

the left upper corner. 

36. The "Fit window" button allows you to zoom automatically the signal 

you are measuring. 

237

37. The "Clean" button can be used to clear the screen at any moment. 

38. Use "Previous" and "Next" buttons to change the visualized angleangle 

plot. 

39. Use "Stop" or "Stop and close" buttons to end the angle-angle plot 


40. If you want to check the results <strong>of</strong> the current measurement before 

proceeding to the next one, press "Open pdf file". 

41. For walking trials, generate a clinical report in just 3 clicks. 

42. In the "SEAG" panel, select "Calibrate". The interface is asking you to 

choose the name <strong>of</strong> the walking trial you want the report. 

43. Click on "Segmentation" button to run the SEAG algorithm for the 

gait event detection. 

44. Click on "Report" to generate the pdf file reports. 

45. If you wish to generate reports which includes the comparison 

between several walking trials, just use the "Comparison" button. 

46. The interface will ask you to select the walking trials that you want to 

compare. 

47. You can compare until 10 walking trials per time. 

48. Congratulations. You can now start using INAIL Manager with 

Outwalk protocol. 

238

239

CHAPTER 5 


UPPER-EXTREMITY THROUGH 

INERTIAL AND MAGNETIC 

MEASUREMENT SYSTEMS 

ABSTRACT 


5.1.1 DEVELOPMENT OF A PROTOCOL FOR THE EVALUATION OF UPPER-EXTREMITY 

KINEMATICS 

5.1.2 APPLICATION SCENARIOS 



PROTOCOLS BASED ON INERTIAL SENSORS 

5.2.1 IDES MANAGER AND ITS USE IN CLINICAL SETTINGS 

5.2.2 IDES MANAGER TUTORIAL 

ABSTRACT 

The aim <strong>of</strong> this chapter is to provide a brief description <strong>of</strong> the upper limb 

functional evaluation protocol <strong>based</strong> on inertial sensors, specifically designed 

for the <strong>analysis</strong> <strong>of</strong> the scapulo-humeral rhythm on patients with shoulder 

pathologies, in clinical settings. 

The chapter provides also the inter-rater reliability study <strong>of</strong> the protocol above 

and its accuracy when comparing it with a scapula locator for static 

measurements <strong>of</strong> the scapula 3D kinematics. 

240


5.1.1 Development <strong>of</strong> a protocol for the evaluation <strong>of</strong> upperextremity 

kinematics 

5.1.1.1 Protocol description 

In [1] iDES, a protocol for Inertial and Magnetic Measuring Systems, which 

allows the easy estimation <strong>of</strong> the elbow, humero-thoracic and scapulo-thoracic 

kinematics was proposed. In particular the protocol was intended for the MT9B 

(Xsens Technologies, NL). 

Thorax, scapula, humerus and forearm were assumed as the segments forming 

the upper-limb. In order to define bone embedded SoRs representative <strong>of</strong> the 

functional anatomy <strong>of</strong> the joints formed by these segments, the robotic 

convention was followed which requires the definition for each segment <strong>of</strong> as 

many SoRs as the number <strong>of</strong> articulations it forms. Since scapulo-thoracic and 

elbow joints are <strong>of</strong> interest, thorax, scapula and forearm form only one 

articulation and had therefore associated only one bone-embedded SoR. The 

humerus, instead, forms two articulations <strong>of</strong> interest, i.e. humero-thoracic and 

elbow: for this reason a system <strong>of</strong> reference was defined for the proximal 

humerus (humero-thoracic joint) and one for the distal humerus (elbow). For 

the thorax, scapula, and proximal humerus, the ISG recommendations [2] were 

adopted. Both H1 and H2 definitions for the proximal humerus were 

implemented. Since the elbow was assumed as a double hinge-joint with nonintersecting 

axes <strong>of</strong> rotation, the SoR <strong>of</strong> forearm and distal humerus were <strong>based</strong> 

on the estimation <strong>of</strong> the functional axes <strong>of</strong> rotation <strong>of</strong> the joint [3]. 

The protocol was developed <strong>based</strong> on the following steps: 

1) SU positioning on the subject. 

One SU is placed on each segment with double-sided tape. For the thorax, the 

SU is positioned on the flat surface <strong>of</strong> the sternum, with the SU z axis exiting 

from the body. For the scapula, the SU x axis is aligned with the cranial edge <strong>of</strong> 

the trigonum spine, in the central third <strong>of</strong> the scapula. For the humerus, the SU 

is just positioned and oriented to minimize the s<strong>of</strong>t-tissue artefact. For the 

forearm, the base <strong>of</strong> the SU is positioned on the distal, flat surface <strong>of</strong> radius and 

241

ulna, with the local z axis exiting from the wrist. 

2) Definition <strong>of</strong> thorax, scapula and proximal humerus bone-embedded SoR by 

means <strong>of</strong> a static acquisition. 

A static acquisition is performed (Figure 1), with the subject standing straight, 

with the arm along the body, perpendicular to the ground. The thorax boneembedded 

SoR is obtained starting from the sagittal plane defined by the 

gravity line and the SU z axis. The scapula SoR is computed from the SU x 

local axis (assumed as the scapula tilting axis) and the gravity line. The 

proximal humerus SoR is assumed to be coincident with that <strong>of</strong> the thorax 

during the static trial. 

3) Definition <strong>of</strong> distal humerus and forearm SoR <strong>based</strong> on elbow axes <strong>of</strong> 

rotation. 

Finally, the subject is asked to perform a pure flexion-extension and a pure 

prono-supination task, thus obtaining an estimation <strong>of</strong> the elbow axes <strong>of</strong> 

rotation. The humerus distal SoR is then defined <strong>based</strong> on the long axis <strong>of</strong> the 

proximal humerus and the flexion-extension axis. The forearm SoR is defined 

<strong>based</strong> on the prono-supination axis and the SU local z axis. The protocol 

presented here appears novel in the literature since none <strong>of</strong> the <strong>protocols</strong> 

developed so far for ISS has addressed the problem <strong>of</strong> measuring the scapula 

kinematics as well as the full 3D joint kinematics <strong>of</strong> the upper-limb. For the 

elbow, the protocol presented in [4] appears close to the present protocol, but its 

performances were conditioned by the wrong assumption <strong>of</strong> a null elbow 

carrying-angle. 

Figure 1 - Static calibration posture with 90° elbow flexed 

242

5.1.1.2 Reliability <strong>of</strong> the protocol 

AMBULATORY MEASUREMENT OF THE 

SCAPULOHUMERAL RHYTHM: INTRA- AND INTER-RATER 

RELIABILITY OF A PROTOCOL BASED ON INERTIAL & 

MAGNETIC SENSORS 

Gar<strong>of</strong>alo P, Cutti AG, Parel I, Fiumana G, Porcellini G, Cappello A 

Proc. Rehab Move, 4th International State‐<strong>of</strong>‐the‐art Congress, 

Rehabilitation: Mobility, Exercise & Sports, 7-9 April, 2009, Vrije 

Universiteit, Amsterdam 

Abstract 

A new protocol has been recently proposed to measure the coordinated 

movement <strong>of</strong> humerus and scapula, through an inertial & magnetic 

measurement system, in ambulatory settings. Since the protocol requires the 

intervention <strong>of</strong> a rater, the aim <strong>of</strong> this study was to assess its intra- and interrater 

reliability. Results for the coefficient <strong>of</strong> multiple correlation showed a 

reliability <strong>of</strong> the protocol ranging from 0.84 to 1, thus supporting its use for 

clinical assessment. 


A clinical parameter which is heavily affected in most <strong>of</strong> shoulder 

musculoskeletal disorders is the scapulohumeral rhythm (SHR). The SHR is the 

coordinated movement between scapula and humerus, when this latter is 

elevated. From the clinical view-point, the SHR is primarily analyzed by 

looking at two angle-angle plots: in the first, the X axis reports the 

humerothoracic elevation and Y the scapulothoracic medio-lateral rotation 

(MELA); in the second, Y reports the scapulothoracic protraction-retraction 

(PRRE). Despite <strong>of</strong> its importance, the measure <strong>of</strong> the SHR has been limited so 

far, also because <strong>of</strong> the lack <strong>of</strong> low-cost, easily-usable, ambulatory 

243

measurement systems. Recently, a protocol has been developed <strong>based</strong> on the 

Xsens inertial & magnetic measurement system (Xsens Technologies, NL), to 

overcome these limitations [1] . 

The Xsens system consists <strong>of</strong> lightweight boxes, called MTx, which comprise a 

3D accelerometer, gyroscope and magnetometer. Through sensor-fusion 

algorithms, the 3D real-time orientation <strong>of</strong> each MTx is known relative to an 

ubiquitous global coordinate system, <strong>based</strong> on the magnetic north and the 

gravity. 

To measure the SHR, the protocol requires to complete the following two 

preliminary steps: 

1. position an MTx sensor on thorax, scapula, and humerus (Fig. 1a): for 

the thorax the sensor is place on the sternum; for the humerus is 

positioned just to minimize the s<strong>of</strong>t tissue artifact; for the scapula, the 

long side <strong>of</strong> the MTx is aligned with the cranial edge <strong>of</strong> the spine, over 

its central third; 

2. execute a static calibration with the subject standing in a pre-defined 

posture: upright position, elbow flexed 90°, humerus perpendicular to 

the ground and in neutral internal-external rotation. 

These steps are required to execute the so-called ―sensor-to-segment 

calibration‖, i.e. to define the anatomical coordinate systems <strong>of</strong> thorax, scapula 

and humerus and relate them to the technical coordinate systems <strong>of</strong> the 

corresponding MTx sensors. 

Both steps require the intervention <strong>of</strong> a rater. In particular, to position the 

scapula sensor the rule has to be followed with care. The rater is also 

responsible to position the subject is the calibration posture, checking in 

particular the orientation <strong>of</strong> the humerus. 

The aim <strong>of</strong> this work was therefore to assess the inter- and inter-rater reliability 

<strong>of</strong> the protocol in measuring the SHR. 

Material and Methods 

A group <strong>of</strong> 20 subjects (25-70 years-old, mean 45) were involved in the 

experiment, together with two raters (A and B) familiar with the protocol. 

Subjects were recovering from different shoulder pathologies, and were 

244

included in the study (after giving their informed consent) if they were able to 

actively and repeatedly elevate the humerus at a minimum <strong>of</strong> 70° in the sagittal 

plane and 45° in the frontal plane. Three measurement sessions were completed 

for each subject with the protocol: two by one rater and one by the other. 

Between acquisitions, the scapula-sensor was removed and reapplied, and the 

static calibration was repeated. Acquisitions were 15 minutes apart, to 

minimize the within-subject biological variability. 

In each acquisition, a total <strong>of</strong> 10 humerus flexion-extensions (FE) and abadductions 

(AA) were measured, but 8+8 were kept for subsequent 

computations. For each movement the SHR waveform was computed and split 

in its upward (flexion, abduction) and downward (extension, adduction) phase 

[5] . 

Before computing the intra- and inter-rater reliability, for each subject we 

checked the intra-subject repeatability <strong>of</strong> the SHR waveforms, by computing 

the adjusted coefficient <strong>of</strong> multiple correlation, in the within-session form 

(CMC 1 ) [5]. Only for those subjects with a very-good repeatability 

(CMC 1 >0.85) we computed the intra- and inter-rater reliability <strong>of</strong> the protocol. 

The intra- and inter-rater reliability was quantified for each <strong>of</strong> the remaining 

subjects by assessing the similarity <strong>of</strong> his/her SHR waveforms with-in and 

between-rater, through the CMC in the between-session form (CMC 2 ) [5], after 

removing the <strong>of</strong>fset between the waveforms <strong>of</strong> the different raters. 

The distribution <strong>of</strong> CMC 2 values over subjects was evaluated with box-andwhisker 

plots, as this was not normal; median and whiskers were extracted; 

values were interpreted as in [5]: 0.75-0.85: good; 0.85-0.95: very-good; 0.95- 

1: excellent. 

Finally, statistically significant differences in the MELA and PRRE ROM 

measured by the raters were searched for through an ANOVA with repeated 

measures. 

Results and Discussion 

Among the 20 subjects, 11 presented a CMC 1 greater than 0.85 in all sessions 

and for all the components <strong>of</strong> the SHR, and were then considered for the intraand 

inter-rater assessment. As discussed in [5], the selection <strong>of</strong> subjects <strong>based</strong> 

on CMC 1 was required, since the CMC 2 can be lowered by a low intra-subject 

repeatability, altering therefore the estimation <strong>of</strong> the intra-rater and inter-rater 

reliability <strong>of</strong> the protocol. 

245

As shown in Figure 1b, higher repeatability was found for the MELA both 

intra- and inter-rater and both for FE and AA, with values within-whiskers 

generally in the excellent range. For the PRRE, values within-whiskers were in 

the very-good to excellent range (1 value out <strong>of</strong> 22 in the good range). The 

intra- and inter-rater reliabilities were comparable for all movements, MELA 

and PRRE. 

No statistically significant differences (p>0.05) were found between the ROMs 

measured by the raters. 

(a) 

(b) 

Figure 1a,b: (a) Xsens sensors set-up. The MTx on the forearm was not used in the present study. 

(b) Distributions <strong>of</strong> CMC 2, intra- and inter-rater, for the FE and the AA task, for MELA and PRRE. 

The CMC 2 values for the upward and downward phase were merged for brevity (22 values are 

reported in each box-plot). 

Conclusion 

Results support the robustness <strong>of</strong> the protocol for longitudinal studies and to 

changes in rater. As a side results, it seems that CMC 1 could be a useful 

parameter to monitor the level <strong>of</strong> mobility restoration <strong>of</strong> subjects, as those 

presenting a lower CMC 1 were actually the subjects clinically evaluated as in 

need for continuing the physical therapy. Further studies are however required 

to draw conclusions. 

246

5.1.1.3 Test <strong>of</strong> the accuracy <strong>of</strong> the protocol 

AMBULATORY MEASUREMENT OF THE 

SCAPULOTHORACIC MOTION: ACCURACY OF A 

PROTOCOL BASED ON INERTIAL AND MAGNETIC 

SENSORS 

Ulrich MJH, van Tuijl EAB, Cutti AG, Gar<strong>of</strong>alo P, Veeger DJ 

Submitted to ISG 2010, 25-27 July 2010, Minnesota (USA) 

Abstract 

This study tests the accuracy <strong>of</strong> a protocol <strong>based</strong> on an inertial and magnetic 

measurement system (IMMS) in quantifying scapulothoracic <strong>motion</strong> (STM). 

Also, a comparison is made between static and dynamic measurements. Fifteen 

subjects performed humerus elevations in the frontal and sagittal plane. 

Outcomes <strong>of</strong> a skin-attached scapula sensor were compared with those <strong>of</strong> a 

sensor on a scapula locator [ 6 ] (Figure 1). Results showed significantly lower 

angles from the skin-attached sensor for medio-lateral rotation and posterioranterior 

tilt in the forward flexion trials, and for medio-lateral rotation in the 

abduction trials. The static-dynamic comparison only showed significant 

differences for posterior-anterior tilt on both humerus movements. The IMMS 

protocol showed to be accurate for protraction-retraction with a trend to be 

more accurate at smaller humerus angles. 

247

Figure 1 – Scapula locator 


An altered scapulothoracic <strong>motion</strong> (STM) can be a sign for disorders <strong>of</strong> the 

glenohumeral joint. This means that attention to altered scapular kinematics is 

important in the clinical evaluation and rehabilitation <strong>of</strong> the shoulder. 

Therefore, accurate measurements <strong>of</strong> the scapular kinematics are <strong>of</strong> great value. 

Unfortunately, in practice, scapular <strong>motion</strong> is difficult to measure. Nowadays a 

detailed 3D <strong>motion</strong> <strong>analysis</strong> can be made using an inertial and magnetic 

measurement system (IMMS). A study <strong>of</strong> Cutti et al. [7] has already shown that 

the protocol for using an IMMS for scapular kinematics [7] has a high intra- and 

interrater reliability. The accuracy <strong>of</strong> the IMMS protocol has, however, not yet 

been tested. The aims <strong>of</strong> this study were (1) to test the accuracy <strong>of</strong> the IMMS 

protocol compared to a scapula locator [3] assumed as the gold-standard, and 

since the scapula locator can only be used statically, (2) to compare the 

differences <strong>of</strong> scapula <strong>motion</strong> between static and dynamic tasks. It was 

hypothesized that no significant differences would have been found between 

results <strong>of</strong> the IMMS protocol and the scapula locator. Also, no differences in 

scapula angles were expected between static and dynamic tasks. 

Materials and methods 

After giving their informed consent, fifteen healthy subjects (9 male and 6 

female, mean age 30.9 years, SD 8.1 years) participated in this study. To test 

the hypotheses, an accuracy protocol and a static-dynamic protocol were 

248

performed. For the accuracy protocol eleven subjects were measured on both 

shoulders. For the static-dynamic protocol twenty-four shoulders were 

measured on fifteen different subjects. 

The MTx IMMS (Xsens Technologies, NL), a commercially available system 

which consists <strong>of</strong> multiple lightweight sensing units (SUs) was used in this 

study. At the start <strong>of</strong> each measurement session, four SUs were attached on the 

skin <strong>of</strong> the subject as in [7]. The anatomical systems <strong>of</strong> reference (SoR) were 

measured during a 10 sec. static trial. 

The scapula locator was used to measure the 3D orientation <strong>of</strong> the scapula 

during different static poses. The legs <strong>of</strong> the scapula locator were designed for 

placing them on the bony landmarks <strong>of</strong> the scapula. An additional SU was 

placed on the scapula locator. The scapula anatomical SoR linked to the 

locator‘s SU was made identical to the anatomical SoR linked to the skinattached 

scapula SU, when the locator was placed on the scapula at 0° <strong>of</strong> 

humerus elevation. Measurements were performed by a paramedical specialist. 

Static data were collected during (1) flexion-extension in the sagittal plane and 

(2) abduction-adduction in the frontal plane. Data were sampled from 0˚ to 140˚ 

in steps <strong>of</strong> 20˚. After each angle, subjects were asked to go back to the resting 

position (circa 0˚ <strong>of</strong> humeral elevation). Scapular orientation was recorded by 

the skin-attached sensor and the sensor on the scapula locator simultaneously, 

by taking one sample at every angle. The measured scapula angles were 

protraction-retraction (PR-RE), posterior-anterior tilting (P-A) and mediolateral 

rotation (ME-LA). 

For gaining the dynamic data, five repetitions <strong>of</strong> FL-EX and AB-AD in 

respectively the sagittal and frontal plane were performed. The subject was 

instructed to make a full RoM and repetitions were done with both arms 

simultaneously. Data from the accuracy protocol were used to make 

comparisons with the dynamic measurements. 

For the statistical <strong>analysis</strong> <strong>of</strong> the accuracy protocol, a 2 x 2 x 8 repeated 

measures ANOVA containing the within factors side (left-right), system (skinattached 

sensor and sensor on scapula locator) and angle (0˚ to 140˚ in steps <strong>of</strong> 

20˚) was used. The dependent variable was the scapula angle. Also, the 

standard error <strong>of</strong> measurement (SEM) was calculated for each humerus angle. 

For the statistical <strong>analysis</strong> <strong>of</strong> the dynamic protocol a 3 x 6 repeated measures 

ANOVA was performed containing the within factors task (static, dynamic 

upwards, dynamic downwards) and angle (20˚ to 120˚ in steps <strong>of</strong> 20˚). A p- 

value ≤ 0.05 was regarded significant. 

249

Results 

For brevity, only the figure <strong>of</strong> the ME-LA movement during the forward 

flexion trials is shown: similar to the other angles, the difference between the 

two systems increased with arm elevation (Figure 2). Scapula locator angles 

were significantly larger than the scapula SU angles except for PR-RE, and P-A 

in the abduction trials (Table 1). Side was never significantly different. 

Comparison with the dynamic humerus elevations (upwards and downwards) 

showed only a significant difference for the P-A angles. For P-A, static angles 

were generally larger than the dynamic angles. 

Figure 2: ME-LA vs. humerus flexion during the static forward flexion task. Scapula locator: 

filled blue line. Scapula SU: dashed green line. SEM: dotted red line. 

Side 

Syste 

m 

Task 

P-A 0.680 0.001 0.001* 

* 

FL-EX ME-LA 0.171 0.00* 0.535 

PR-RE 0.774 0.125 0.385 

P-A 0.454 0.110 0.00* 

AB-AD ME-LA 0.348 0.00* 0.479 

PR-RE 0.361 0.669 0.053 

Table 1: ANOVA p-values <strong>of</strong> static accuracy trials (side, system) and static-dynamic comparison 

(task). *significant with p≤0.05 

250

Discussion and conclusions 

The first hypothesis stated that no significant differences would be found 

between results <strong>of</strong> the IMMS protocol and the scapula locator. This could not 

be confirmed entirely. The IMMS protocol proved to be accurate for PR-RE 

and P-A in the abduction trials and PR-RE in the forward flexion trials. The 

smaller the humerus angles, the more accurate the protocol seems to be. 

Differences between the sensor on the scapula locator and the skin-attached 

sensor in this study were larger compared to the differences between the 

scapula locator and the acromion marker cluster as used in the study <strong>of</strong> van 

Andel et al. [8] . As in their results, this study shows a general underestimation 

<strong>of</strong> the skin-attached sensor compared to the scapula locator derived angles. The 

scapula SU measurements seem clinically acceptable below 100 ˚, since up to 

there the SEM

5.1.2 Application scenarios 

During the validation <strong>of</strong> iDES protocol several scenarios <strong>of</strong> application were 

studied. Among these, an example <strong>of</strong> application <strong>of</strong> iDES on a subject who 

recently underwent shoulder surgery is shown in Figure 3, during a 

rehabilitation session. 

iDES protocol was applied to the impaired shoulder in a novel scenario for 

<strong>motion</strong> <strong>analysis</strong>, which includes the intervention <strong>of</strong> the physiotherapist 

performing the rehabilitation session and the real-time outcome from iDES 

using INAIL Manager, adopted as direct feedback both for the patient and the 

operator. 

Figure 3 – Active mobilization during a rehabilitation session. In the background, the visual 

feedback provided by INAIL Manager 

The movement <strong>of</strong> forward flexion <strong>of</strong> the shoulder during three different 

sessions were compared: before, during, and after the mobilization by the 

physiotherapist. 

Figure 4 shows the final report provided by INAIL Manager, in terms <strong>of</strong> 

scapulo-humeral rhythm. In particular, the correlation between the mediolateral 

rotation <strong>of</strong> the scapulo-thoracic joint and the flexion-extension <strong>of</strong> the 

252

humero-thoracic joint are visualized. 

When comparing the pattern <strong>of</strong> the subject without the help <strong>of</strong> the 

physiotherapist (red pattern) and during the active mobilization (green pattern) 

we can see that they are substantially different, being the help <strong>of</strong> the 

physiotherapist useful for avoiding compensation strategies. 

When the patient is free to continue to perform the movement without any help, 

he is still able to perform the movement correctly (blue pattern). 

Further studies are required to understand how visual feedback can improve or 

modify the scapulo-humeral rhythm exercises while adopting iDES protocol. 

However, results show that the measurement <strong>of</strong> the scapulo-humeral rhythm 

while actual mobilization in clinical settings is promising, being iDES protocol 

versatile and quick to apply. 

Figure 4 – Scapulo-humeral rhythm comparison between the different phases <strong>of</strong> the rehabilitation 

session 

253





2. Helm FVD. A standardized protocol for <strong>motion</strong> recordings <strong>of</strong> the shoulder. 

Proceedings <strong>of</strong> the First conference <strong>of</strong> the International Shoulder Group. 

3. Cutti AG, Gar<strong>of</strong>alo P, Davalli A, Cappello A. How accurate is the estimation 

<strong>of</strong> elbow kinematics using ISB recommended joint coordinate systems Gait & 

Posture. 2006;24:S36-S37. 

4. Luinge HJ, Veltink PH, Baten CTM. Ambulatory measurement <strong>of</strong> arm 

orientation. J Biomech. 2007;40(1):78-85. 




2009;47(5):475-486. 

6. Price CI, Rodgers H, Franklin P, Curless RH, Johnson GR. Glenohumeral 

subluxation, scapula resting position, and scapula rotation after stroke: A 

noninvasive evaluation. Archives <strong>of</strong> Physical Medicine and Rehabilitation. 

2001;82(7):955-960. 

7. Cutti A, Gar<strong>of</strong>alo P, Parel I, Fiumana G, Porcellini G. Intra- and inter-rater 

reliability <strong>of</strong> the scapulohumeral rhythm measured by a protocol <strong>based</strong> on 

inertial and magnetic sensors. Gait & Posture. 30(Suppl. 1):17. 

8. van Andel CJ, Wolterbeek N, Doorenbosch CAM, Veeger DHEJ, Harlaar J. 

Complete 3D kinematics <strong>of</strong> upper extremity functional tasks. Gait Posture. 

2008;27(1):120-127. 

9. Fayad F, H<strong>of</strong>fmann G, Hanneton S, et al. 3-D scapular kinematics during 

arm elevation: Effect <strong>of</strong> <strong>motion</strong> velocity. Clinical Biomechanics. 

2006;21(9):932-941. 

254


SOFTWARE FOR THE PROTOCOLS BASED ON 


The following is a description <strong>of</strong> the typical use <strong>of</strong> INAIL Manager s<strong>of</strong>tware 

for the application <strong>of</strong> the iDES protocol described in the previous section. A 

Tutorial is also provided. 

255

5.2.1 iDES Manager and its use in clinical settings 

Set-up – Positioning <strong>of</strong> the MTx units 

Positioning <strong>of</strong> the MTx units can be obtained by means <strong>of</strong> elastic bands, Co- 

Plus Cohesive bandage (Phoenix Healthcare) and double sided-tape around 

each MTx. For body segments like humerus and forearm, special MVN straps 

(Xsens Technologies B.V.), shown in Figure 5 can be adopted which allow to 

be easily attached and detached during the experiments. However, they must be 

correctly placed depending on the muscle tone <strong>of</strong> the subject, avoiding that they 

can slip away during the movement. The positioning is basically very similar to 

what can be obtained using the Co-plus, but the thickness <strong>of</strong> the straps is 

greater than the Co-plus one and it could create an obstacle to the movement. 

After the use <strong>of</strong> the streps, they should be washed in cold water, or possibly 30 

degrees (also washing machine). Avoid to use spin. 

Figure 5 – MVN straps 

256

The positioning over the thorax, scapula and forearm segments are the most 

critical. iDES protocol indicates that the rear <strong>of</strong> the MTx units (Figure 6) must 

be attached to the body segment. 

Figure 6 – x, y, z axes <strong>of</strong> the MTx sensing unit 

The sensing unit over the thorax segment must be positioned so that the x axis 

points cranially and the z axis points out from the thorax (Figure 7). 

In the case <strong>of</strong> the presence <strong>of</strong> hair over the sternum area, it is also possible to 

position the MTx unit over the C7 dorsal area. However, in order to be 

consistent with the protocol, the x axis is still pointing cranially while the z axis 

must point within the thorax. 

This positioning is valid both for right or left limb measuring. 

Figure 7 – Positioning <strong>of</strong> the MTx unit over the sternum 

257

The sensing unit over the scapula (Figure 8) must be positioned so that the x 

axis <strong>of</strong> the MTx points laterally and is aligned to the middle part <strong>of</strong> the scapula 

spina. Both in the case <strong>of</strong> left or right limb measurement, the cable connectors 

<strong>of</strong> the MTx over scapula point medially. 

Figure 8 - Positioning <strong>of</strong> the MTx unit over the scapula 

The sensing unit over the forearm (Figure 9) must be positioned so that the x 

axis points towards the hand, and aligned with the long axis <strong>of</strong> the forearm. 

Therefore the connectors point towards the elbow. This positioning is valid 

both for right or left limb measurements. 

258

Figure 9- Positioning <strong>of</strong> the MTx unit over the forearm 

For all the MTx units, except for the units over the scapula and the forearm, 

the measurement can be mainly influenced by the way in which the calibration 

procedures are performed. For the unit over the scapula, being it aligned to the 

bone, the positioning is the most critical part. In the previous section the 

reliability in the positioning <strong>of</strong> this sensing unit was assessed. 

How iDES Manager works 

In general, the s<strong>of</strong>tware allows to comunicate with the hardware by two 

different modalities: 

1) Measurement modality 

With this modality the system in active in background and each MTx 

communicates with the Xbus Master sending all the raw data from the inertial 

and magnetic sensors. With this modality, the s<strong>of</strong>tware is ready for asking 

orientation data for each MTx unit and apply iDES protocol, showing the result 

in real-time. 

With this modality the user can only modify the elements regarding the type <strong>of</strong> 

real-time visualization wanted. 

259

This is also the modality to adopt during the initialization <strong>of</strong> the system. 

As long as this modality is active, the s<strong>of</strong>tware records the raw data from the 

sensing units even if the user does not choose to plot any data. 

This modality can be activated through the ―Start Measuring‖ button (Figure 

10). 

Figure 10 – iDES Manager acquisition interface 

2) Configuration modality 

With this configuration the system is in stand-by, waiting for receiving 

commands <strong>of</strong> configuration (e.g. sampling frequency, kalman-filter scenario 

etc..). With this modality the s<strong>of</strong>tware does not record any kind <strong>of</strong> data because 

no data is available. 

This modality can be activated through the ―Stop Measuring‖ button (Figure 

11). 

260

Inizialization <strong>of</strong> the system 

In order to get an accurate kinematic data the system needs to be inizialized. 

In order to do this, it is sufficient that the system remains in static conditions 

during the measuring modality, for at least 60 seconds. 

However, the following procedure is suggested: 

a) start the measuring modality, leaving the sensign units in static condition at 

least for 40 seconds, that is the subject must stand still; 

b) ask the subject to perform slow and not complex movements for other 40 

seconds 

c) ask the subject to return to a static position for other 10 seconds 

Note that during this procedure the user can visualize 3D orientation data in 

real-time in order to verify the stability <strong>of</strong> the signals. 

d) now the user has two possibilities, (1) to leave the system in measurement 

modality and proceed with the execution <strong>of</strong> the entire protocol or (2) store the 

state <strong>of</strong> the system in order to avoid to repeat the executation <strong>of</strong> the 

initialization procedure when some problems occur 

In the future, this procedure will become automatic, simplifying the operations 

by the user. 

261

5.2.2 iDES Manager Tutorial 

1. Welcome to INAIL Manager. This tutorial will guide you through the 

use <strong>of</strong> INAIL Manager and iDES Protocol in particular. 

2. First, press "Scan" to connect to your system and checking for the 

MTx units available. 

3. Now, select the corresponding body segment for each MTx unit. Press 

"Next" to go to the main acquisition interface. 

4. Press "Change Folder" to select the main folder in which all patients 

data will be stored. 

5. Press "New patient" to create a new patient in the database. 

6. If you want to select a patient previously acquired, press "Find other 

patient" button. 

7. Before proceeding with the protocol steps, let's have a brief 

introduction about how to control the measurement through the 

interface buttons. 

8. In the lower left cornet, the "Control Panel" allows you to control the 

connection between the system and the data visualization. 

9. "Start Measuring" change the modality <strong>of</strong> the system into 

Measurement Mode, while with "Stop Measuring" the system enters 

the Configuration Mode. 

10. In the upper right corner, the "Xbus Master Settings" allow you to 

control the acquisition parameters. These settings will be applied only 

if the "Start Measuring" button is pressed. 

11. "Start plotting" and "Stop Plotting" control the data visualization. Joint 

angles are displayed in terms <strong>of</strong> the three angle components. 

262

12. Press 'n' to change the visualized joint. 

13. Press 'z' to zoom in the signals. Press 'z' again to restore the normal 


14. "Save filter state" can be pressed before or during the data 

visualization to store the state <strong>of</strong> the system. The state will be restored 

everytime "Start Measuring" button is pressed. 

15. "Rescan" button can be used to go back to the previous interface, for 

changing the body segments order or to start again the communication 

with the system when it was accidentally lost. 

16. If you just entered the acquisition interface you must perform the 

initialization <strong>of</strong> the system before proceeding. 

17. Press "Start Measuring". The system is now in Measurement Mode 

and all the settings you selected through the interface were applied to 

the system. 

18. Wait about 1 minute. During this period, the subject should stand still 

for 40 seconds and than move around for some seconds and stand still 

again. 

19. If you do not want to save the state <strong>of</strong> your system, just leave the 

system in Measurement Mode. 

20. The first step <strong>of</strong> iDES protocol is to perform a static calibration trial. 

21. Press "Start Plotting" to visualize real-time joint angles during the 

static calibration posture. 

22. Wait at least 8 seconds before pressing "Stop plotting" or the X in the 

upper corner. If the static trial was performed correctly, answer 'yes' 

for saving data and proceeding with the next step. 

23. The second step <strong>of</strong> iDES protocol is to perform the estimation <strong>of</strong> the 

flexion-extension functional axis <strong>of</strong> the elbow. 

263

24. Press again "Start Plotting" to visualize the elbow joint angles. Ask the 

subject to perform some consecutive pure elbow flexion-extension, 

with some short breaks between the flexions and the extensions. 

25. Then close the visualization. The system will indicate the "dispersion 

angle" parameter. Check if your results is around 10 degree and then 

save your data. 

26. If you want, you can perform, visualize and save more trials and then 

in the end choose the best one. 

27. The third step <strong>of</strong> iDES protocol is to perform the estimation <strong>of</strong> the 

prono-supination functional axis <strong>of</strong> the elbow. 

28. Follow exactly the same instructions as before, asking to the subject to 

perform some consecutive pure elbow prono-supination tasks, with 

some short breaks between the pronation and the supination. 

29. You are now ready to measure every kind <strong>of</strong> upper limb activity. Just 

enter the name <strong>of</strong> the trial and press "Start Plotting" to get real-time 

3D joint angles. 

30. If you want to visualize your data using angle-angle plots, first click 

on "Enable coordination plots", 

31. choose the kind <strong>of</strong> angle-angle plot do you prefer, 

32. now press "Start Plotting". 

33. In the angle-angle plot interface, the acquisition time is displayed in 

the left upper corner. 

34. The "Fit window" button allows you to zoom automatically the signal 

you are measuring. 

35. The "Clean" button can be used to clear the screen at any moment. 

264

36. Use "Previous" and "Next" buttons to change the visualized angleangle 

plot. 

37. Use "Stop" or "Stop and close" buttons to end the angle-angle plot 


38. If you want to check the results <strong>of</strong> the current measurement before 

proceeding to the next one, press "Open pdf file". 

39. Congratulations. You are now able to use INAIL Manager with iDES 

Protocol. 

265

266

CHAPTER 6 

A NEW ALGORITHM FOR THE 

APPLICATION ON AMPUTEES OF THE 

LOWER AND UPPER-EXTREMITY 

PROTOCOLS BASED ON INERTIAL 

SENSORS 

ABSTRACT 

6.1 INTRODUCTION 

6.2 KIC (KINEMATIC COUPLING) ALGORITHM 

6.3 INTERFACING KIC ALGORITHM WITH UPPER AND LOWER-EXTREMITY 

PROTOCOLS 


ABSTRACT 

The aim <strong>of</strong> this chapter is to give a brief description about the application <strong>of</strong> 

inertial and magnetic measurement units for the kinematic <strong>analysis</strong> <strong>of</strong> upper 

and lower limb amputees. First a brief introduction about how to use 

information from magnetometers together with inertial sensors in the sensing 

fusion algorithms is provided. Then, a new Kalman filter-<strong>based</strong> algorithm is 

proposed for enabling the <strong>analysis</strong> without the use <strong>of</strong> magnetometers for the 

estimation <strong>of</strong> the 3D joint orientation. 

The application <strong>of</strong> the new method and how it can be interfaced with the upper 

and lower limb <strong>protocols</strong> described in the previous chapters are presented. 

267

6.1 Introduction 

As previously discussed in Chapter 1, <strong>motion</strong> capture systems <strong>based</strong> on 

accelerometers and gyroscopes can be used for the estimation <strong>of</strong> 3D orientation 

and position. However, drift errors occur during double integration <strong>of</strong> 

acceleration for the estimation <strong>of</strong> 3D position and during single integration <strong>of</strong> 

angular velocity for the estimation <strong>of</strong> 3D orientation. 

Errors in the estimation <strong>of</strong> the bias <strong>of</strong> the gyroscopes can be corrected by means 

<strong>of</strong> Kalman filter-<strong>based</strong> algorithms, but the estimation <strong>of</strong> the absolute heading 

angle <strong>of</strong> the inertial platform, i.e. the angle <strong>of</strong> rotation <strong>of</strong> the inertial platform 

around the vertical direction, expressed in a global coordinate system, is still 

―not observable‖ if only the information from accelerometers and gyroscopes 

are fused together. ―Not observable‖ means that the heading angle is a part <strong>of</strong> 

the predictive-corrective system which no correction can be applied to. 

Therefore it can be estimated, but that value is reliable only for a short period. 

Magnetic sensors, which are also adopted as basic working principle in 

electromagnetic devices and, more specifically, information about the earth 

magnetic field, can be used in order to make the heading ―observable‖ within 

the system, as described in [1]. 

Once the heading becomes observable, 3D orientation can be estimated within 

1 degree <strong>of</strong> static accuracy in environments with a homogeneous magnetic field 

(Xsens MTx manual) and 2 degrees RMS in terms <strong>of</strong> dynamic accuracy. This 

may change according to the type <strong>of</strong> movement performed. The ideal condition 

for getting the best estimate <strong>of</strong> the 3D orientation is when there are no magnetic 

distortions at all, which means that the earth magnetic field is measured by the 

system. 

When there are magnetic distortions but the resulting magnetic field is 

homogeneous, the system can be ―trained‖ to the new magnetic field adopting 

several techniques, like the Magnetic Field Mapping (Xsens manual), as 

described in [2]. 

From the above it comes clear that for non homogeneous magnetic fields the 

system can present limitations as the electromagnetic devices do. 

Many different sources can generate non homogeneous magnetic field. 

Buildings, rooms, laboratories, are normally affected by magnetic field 

distortions which change depending on the area in which we move [3]. 

Force plates are one <strong>of</strong> the potential causes <strong>of</strong> magnetic distortions inside <strong>of</strong> the 

laboratories. Moreover, some prosthetic devices contain ferromagnetic objects 

which can change the magnetic field around it, and this can be critical at 

268

different levels. Ankle prostheses may contain metal screws which can provide 

the magnetic field to change into a new, although homogeneous, magnetic 

field. Microprocessor-controlled prostheses with no active motors may 

influence the magnetic field during dynamic tasks. At the highest level we 

found active prostheses which include active electric motors. 

In this cases, different methods can be adopted, depending on the context and 

the specific application. 

One method consists in enhancing the Kalman filter-<strong>based</strong> algorithm in order to 

improve its robustness to the variation <strong>of</strong> the magnetic field. This method is 

<strong>based</strong> on various assumptions made on the external magnetic field. Of course, 

not every kind <strong>of</strong> variation can be corrected for. The above method includes 

assumptions made on the measured signals, like magnetic field. This method 

can be useful when magnetic field is influenced by objects which are passed by 

the subject during the measurements. 

Other methods for the estimation <strong>of</strong> 3D orientation are <strong>based</strong> on assumptions 

made on the anatomy or kinematics <strong>of</strong> body segments. Some methods were 

adopted in literature which considered assumptions on the body segments 

lengths around a specific joint. This information is used for correcting the 

estimation provided by the Kalman filter-<strong>based</strong> algorithm. This is an example 

<strong>of</strong> ―biomechanical constraints‖ used as additional information to the overall 

system. 

Another method which will be presented in the next section, named KiC 

(Kinematic coupling), developed by Xsens Technologies B.V., considers the 

assumption that for a specific joint, two body segments are assumed as the 

proximal and distal segment with a common joint, therefore moving together 

with the joint in such a configuration. The way in which this information is 

used and the way in which this configuration is modelled (e.g. whether the joint 

is considered to have some laxity or the number <strong>of</strong> degrees <strong>of</strong> freedom) play an 

important role in the 3D kinematics estimation. 

It is important to notice that this last method consists in providing the Kalman 

filter-<strong>based</strong> algorithm with additional information which is fused together with 

information from the inertial sensors. 

When this method is adopted inside <strong>of</strong> the Kalman-filter <strong>based</strong> algorithm, the 

relative heading becomes observable. In other terms, provided that a proximal 

and distal joint are connected through a joint, the relative orientation <strong>of</strong> the 

distal segment with respect to the proximal one can be calculated without the 

use <strong>of</strong> magnetometers, i.e. the electromagnetic distortions in the environment 

do not influence the estimation <strong>of</strong> the 3D joint kinematics. 

269

KiC algorithm uses a priori information about the distances between the 

sensing units positioned over the proximal and distal segments and the adjacent 

joint for the calculation <strong>of</strong> the joint 3D kinematics. As long as the joint is 

moving, the relative heading is observable. An important aspect to consider is 

that when no information about the magnetic field are used in the algorithm, the 

absolute heading is not observable. This means that calibration procedures 

including the sensor-to-segment calibration when an accurate estimation <strong>of</strong> the 

absolute orientation <strong>of</strong> the body segments is needed, KiC must be adopted 

using additional methods or specific procedures. 

For this aspect, together with an evaluation <strong>of</strong> the robustness <strong>of</strong> the algorithm to 

variability in the estimation <strong>of</strong> the distances required as input, further studies 

are required. 

Application <strong>of</strong> KiC on a microprocessor-controlled knee prosthesis was 

provided in Chapter 4. 

270

6.2 KiC (Kinematic Coupling) algorithm 

A METHOD FOR GAIT ANALYSIS USING INERTIAL 

SENSORS 

Schipper L, Roetenberg D, Gar<strong>of</strong>alo P, Cutti AG, Luinge HJ 

Accepted at JEGM 2010, Miami (US) 


Inertial sensors have been proposed and successfully applied for ambulatory 

gait <strong>analysis</strong>. 

Although inclination can be measured with high accuracy using gyroscopes and 

accelerometers, heading tracking can be challenging for some applications. A 

common way to estimate absolute heading is by adding complementary 

sensors, usually magnetometers. Locations in which gait <strong>analysis</strong> is performed 

do not always have a homogenous magnetic field which in turn leads to 

incorrect heading estimates [1]. However, because only the relative orientation 

between two segments is required to compute the joint angle, there is no need 

for an absolute heading reference. This abstract presents a novel method for 

stable and accurate tracking <strong>of</strong> 3D joint angles that does not reference to the 

local magnetic field. 

Clinical Significance 

Gait <strong>analysis</strong> using miniature inertial sensors can accurately be measured using 

the so-called Kinematic Coupling (KiC) algorithm. It does not use 

magnetometers or the local magnetic field to stabilize heading. Therefore, this 

method is suitable to determine the gait pattern for many consecutive strides in 

any daily live environment without the need for a fixed infra-structure. 

Methods 

271

ANKLE 

KNEE 

The KiC algorithm estimates the orientation <strong>of</strong> three adjacent segments with the 

joints modeled as ball-and-socket joints containing some laxity. The 

gyroscopes are used to predict the change in angle <strong>of</strong> each segment. The gravity 

vector measured with the accelerometers is used for inclination estimation <strong>of</strong> 

each segment. With known distances between sensors and joints, the relative 

heading is estimated by assuming that the sensors attached to the segments 

measure the same joint acceleration when the joint is subject to acceleration. 

The lower limb kinematics <strong>of</strong> the unimpaired limb <strong>of</strong> a transfemoral amputee 

during walking were measured with Xsens sensors and with an optical system 

Vicon (trial 1). For details about the setup and how the two kinematics were 

compared, see [2]. With this set-up, the s<strong>of</strong>t tissue artefacts are equal for both 

systems. A second trial was recorded with Xsens sensors only, while the 

subject was walking in a straight line for more than 20 successive strides. The 

Cast protocol [2] was applied as calibration method to align sensors to 

segments. 

Results 

The knee and ankle joint angles for trial 1 and 2 are shown in Figure 1 and 2. 

The mean RMS difference between KiC and the optical system over all joint 

angles for 10 strides is 2.0 degrees. The 20 successive strides show a high 

repeatability and no drift. 

70 

50 

F(+)E(-) (deg) 

25 

15 

Ab(+)Ad(-) (deg) 

5 

-5 

I(+)E(-) (deg) 

5 

-10 

15 

0 

-20 

-40 

20 60 100 

20 60 100 

Stride (%) 

0 

-10 

5 

-5 

-15 

-20 

20 60 100 

20 60 100 

Stride (%) 

-15 

-25 

35 

20 

10 

5 

20 60 100 

20 60 100 

Stride (%) 

Figure 1: 3D knee and ankle joint angles for one typical stride during walking, in black the KiC 

algorithm and in gray the optical system. The mean RMS difference between KiC and the optical 

system over all joint angles for 10 strides is 2.0 degrees which is in the order <strong>of</strong> the accuracy <strong>of</strong> the 

optical system. 

272

ANKLE 

KNEE 

70 

50 

F(+)E(-) (deg) 

25 

15 

Ab(+)Ad(-) (deg) 

5 

-5 

I(+)E(-) (deg) 

5 

-10 

15 

0 

-20 

-40 

20 60 100 

20 60 100 

Stride (%) 

0 

-10 

5 

-5 

-15 

-20 

20 60 100 

20 60 100 

Stride (%) 

-15 

-25 

35 

20 

10 

5 

20 60 100 

20 60 100 

Stride (%) 

Figure 2: 3D knee and ankle joint angles for 20 consecutive strides during walking using KiC show 

stable and consistent tracking <strong>of</strong> joint angles. Variations over steps can be explained by small 

stride-to-stride variation and do not contain outliers. 


From Figure 1, it can be concluded that the KiC algorithm is an accurate 

method for ambulatory gait <strong>analysis</strong>. Based on the used measurement set-up, 

the observed differences are in the order <strong>of</strong> the accuracy <strong>of</strong> the optical system. 

Furthermore, the computed joint angles show high repeatability as can be seen 

in Figure 2. The stride-to-stride variation can be explained by the natural 

variation within a normal gait pattern. 

273

6.3 Interfacing KiC algorithm with upper and lower-extremity 

<strong>protocols</strong> 

1. Introduction 

INAIL Prostheses Centre was provided with the KiC (Kinematic Coupling) 

algorithm as an alternative to the XKF3 Kalman filter provided with the Xbus 

kit system <strong>based</strong> on MTx (Xsens Technologies B.V.) for the estimation <strong>of</strong> MTx 

3D orientation. The s<strong>of</strong>tware was provided as a Windows stand-alone 

application which can be run both through command prompt and Matlab 

command window. Both as 32 or 64 bit versions are available. 

The s<strong>of</strong>tware can be considered as a black box with several inputs and outputs. 

Being INAIL Manager adopted for using XKF3 for the application <strong>of</strong> iDES and 

Outwalk <strong>protocols</strong>, the aim <strong>of</strong> this section is to provide preliminary indications 

about how KiC executable can be integrated with INAIL Manager. 

2. Methods 

2.1 Characteristics <strong>of</strong> KiC executable 

Firstly we summarize some characteristics <strong>of</strong> the current KiC algorithm which 

can be considered as important specifications for the integration with INAIL 

Manager: 

 

 

 

 

The current version <strong>of</strong> KiC executable (v1.0.0) contains the algorithm 

which estimates the 3D orientation <strong>of</strong> three MTx units inside <strong>of</strong> a 

kinematic chain formed by three segments and two joints (e.g. thigh, 

shank and foot segments with knee and ankle as joints) 

KiC runs on 3 segments per time, which means that only one limb can 

be processed 

Differently from XKF3, only two different scenarios are selectable in 

the current KiC 

Currently KiC executable cannot be adopted for real-time usage. KiC 

executable can be applied directly on a raw data .mtb or .xm file 

provided as output through INAIL Manager 

274

2.2 Assumptions 

As a preliminary document, some assumptions are made: 

 

 

 

 

 

When the goal <strong>of</strong> the user is to use KiC executable inside <strong>of</strong> a specific 

protocol (Outwalk or iDES), we assume that every kind <strong>of</strong> data given 

as an input to KiC was previously processed using XKF3 algorithm. 

This is because information about events like the pressing <strong>of</strong> the Start 

Measuring button and Start Plotting are only available inside <strong>of</strong> the 

.mat files given as output <strong>of</strong> INAIL Manager using XKF3. The time 

events information are available in the ―protocol free‖ module in 

INAIL Manager as well. 

Note that if the user does not apply a specific protocol to the data or 

the user manually takes notes about the time events, integration <strong>of</strong> KiC 

with INAIL Manager is not necessary, being the collection <strong>of</strong> the raw 

data possible through other s<strong>of</strong>twares. 

We assume that the first protocol taking advantage <strong>of</strong> KiC executable 

will be Outwalk. However, the integration <strong>of</strong> KiC with INAIL 

Manager will take into account the fact that other <strong>protocols</strong> can be 

adopted. 

For applying Outwalk protocol to output data from KiC executable, 

we assume that at least the static calibration step was already 

processed using XKF3. In fact, while preliminary tests on processing 

knee functional axis calibration through KiC were performed, no test 

results can be provided for the static calibration yet. 

Information about distances between the origin <strong>of</strong> the MTx embedded 

frame and the joint centers can be calculated using several methods 

(optoelectronic systems, rulers, pictures etc…). For now, we will not 

take this aspect into account. 

275

2.3 Proposal <strong>of</strong> a step by step procedure for using KiC 

through INAIL Manager 

1. The user collected kinematic data through INAIL Manager using 

Outwalk Manager. 

Settings 

2. The user selects the filter scenario to be used by KiC, through a popup 

menu in INAIL Manager. 

3. The user clicks on the ―Apply KiC‖ button on the interface and selects 

the raw data .mtb or .xm file containing the data acquired previously. 

4. For the selected raw data file the interface looks for information about 

which segments are available in the data. 

5. The user is asked to select which limb needs to be processed through 

KiC and the scenario previously selected. 

6. The interface extracts the MTx device IDs information from the limb 

selected by the user. 

7. A new window now appears asking the user to provide information 

about the distances between the MTx units and the adjacent joint. 

Running KiC 

8. The interface automatically runs KiC executable on the selected raw 

data file, using as input the device IDs, the selected scenario and the 

information about the distances. In this phase all the outputs from KiC 

executable will be stored into a temporary location. 

Post processing 

9. The interface now automatically looks for all the calibration .mat files 

and dynamic mat file in which the information about the time events 

along the raw data mat file were previously saved. 

276

10. The interface shows all the trials available inside <strong>of</strong> the raw data file 

(hence inside <strong>of</strong> the KiC output), starting from the .mat files available. 

11. Before the user decides which protocol should be applied to the data, 

all the output from KiC is automatically converted into the same 

format available in INAIL Manager. 

12. The user is allowed to choose among one <strong>of</strong> the following: 

a. Starting from the calibration data already available and 

previously processed using XKF3, to apply Outwalk protocol 

on a specific trial available inside <strong>of</strong> the output <strong>of</strong> KiC; 

b. To apply Outwalk protocol entirely, including static 

calibration step, functional calibration step and dynamic step 

on the corresponding parts inside <strong>of</strong> the output <strong>of</strong> KiC; 

c. To apply a different protocol than Outwalk on a specific part 

<strong>of</strong> the output <strong>of</strong> KiC (e.g. to apply specific sensor-to-segment 

calibration matrices to the orientation data) 10 

13. The protocol selected in 12 is now applied to KiC output for the 

selected limb and new .mat files, with the same structure as the one in 

INAIL Manager, is created. 

14. A new report (.ps <strong>of</strong> .pdf format) is now created as the one always 

produced by INAIL Manager at the end <strong>of</strong> the processing 11 . 

15. The user is now asked to apply the entire procedure (from 5 to 13) on 

the other limb available, to repeat all the procedure (e.g. selecting a 

10 Note that this is the case in which KiC is adopted for accuracy tests, when 

comparing two different measurement systems or two different Kalman filter<strong>based</strong> 

algorithms. 

11 Note that at the end <strong>of</strong> this procedure the Outwalk report generation can be 

normally performed on the new .mat file created. Similarly for the comparison 

report generation. 

277

different scenario or different values <strong>of</strong> distances) or to end the 

application <strong>of</strong> KiC. 

2.4 Proposal <strong>of</strong> how KiC should be integrated with INAIL Manager for 

satisfying the step by step procedure 

Figure 2 shows the main modules which allow KiC to be integrated with 

INAIL Manager. A specific role is assigned to each module. 

Figure 2 – Modules for the integration <strong>of</strong> KiC with INAIL Manager 

Module A 

It realizes the connection between the INAIL Manager output format and the 

KiC executable input format. Basically it looks for the necessary information 

for KiC and it converts it as readable information by KiC. Through an external 

window, this module also manages the inputs about the distances between the 

MTxs and the adjacent joint. This module covers steps 2-7. Potentially, it could 

cover also step 15, being Module A a sort <strong>of</strong> ―main function‖. 

Module B 

It realizes the recognition <strong>of</strong> the available trials along the raw data file. 

Basically it looks for information about the events like Start Measuring and 

Start Plotting buttons pressing, available inside <strong>of</strong> the .mat files previously 

obtained through INAIL Manager using XKF3 12 . Then, the KiC output is split 

12 Note that the information about the ―stops‖ can be easily obtained through 

the start events and the length <strong>of</strong> the data available. 

278

into several parts, depending on the trials previously recognized. This module 

covers steps 9-10. 

Module C 

It covers step 11. The conversion is done on all the data available as KiC 

output. 

Module D 

It covers step 12, in which the user selects how to treat the available data. 

Basically this module will work as a ―selector‖ <strong>of</strong> <strong>protocols</strong> and a preliminary 

preparation <strong>of</strong> the data to be processed will be performed, which then will be 

treated by Module E. 

Module E 

It realizes the application <strong>of</strong> the protocol selected by the user. The input <strong>of</strong> this 

module is the output <strong>of</strong> KiC already converted by module C. The information 

about which protocol should be launched will come from Module D. Basically 

this module substitutes the structure containing the sensor unit orientation 

provided by XKF3 with the one provided by KiC. Then the new protocol will 

be applied starting from this information. It covers steps 13 and 14. 

3. Discussion 

Discussion about the proposal 

The proposal was designed so that code previously created for INAIL Manager 

could be used and new features to INAIL Manager can be added at the same 

time. 

Module A can be easily created, being the link to data already available in mat 

files. The scenario selection, instead, should be done by reading the current 

values on the INAIL Manager interface. Information about the body segments 

available and the corresponding device IDs is already among the .txt 

parameters files in the INAIL Manager output. Some adjustments are probably 

needed for getting the right numeric format. Note that this module could be 

created as an independent code/interface, so that KiC executable could be used 

also without running INAIL Manager (e.g. collecting raw data through MT 

Manager and then running KiC on a robot or prosthesis). 

Module B is currently not supported by any available code, but its realization 

279

should not be difficult. The way in which the available data is presented to the 

user can be a point <strong>of</strong> discussion. 

Module C is partially available through codes previously created for converting 

Outwalk Manager output to a new data structure more readable and flexible. 

This was done in order to easily apply Outwalk protocol in <strong>of</strong>fline mode. A 

document about this point is available. From the point <strong>of</strong> view <strong>of</strong> the 

implementation, we should decide either to use the new data structure to make 

all INAIL Manager more flexible or to create a converter from KiC output to 

INAIL Manager old data structure currently available. 

Module E is partially available through the <strong>of</strong>fline code <strong>of</strong> Outwalk protocol, in 

which the user can already choose which step <strong>of</strong> the protocol must be applied. 

For this reason, module E could be created as an independent module, which 

allows the user to apply different <strong>protocols</strong> even if INAIL Manager is not 

running. Some adjustments have to be done when dealing with iDES protocol 

<strong>of</strong>fline codes. 

Module D is the link between the choices <strong>of</strong> the user and the application <strong>of</strong> the 

<strong>protocols</strong>. The way in which this can be done is an implementation issue. 

Module E and D could be fused together or module E could split on several 

modules, each one realizing a different protocol. 

The proposal already contains information about what should be created, and 

how the modules should be connected each other. The way in which this can be 

done needs to be discussed because it depends on which future changes or 

upgrades in INAIL Manager need to be done. For example, the creation <strong>of</strong> all 

the modules should take into account the future use <strong>of</strong> KiC in real time or the 

creation <strong>of</strong> a complete version <strong>of</strong> INAIL Manager <strong>of</strong>fline version. 

280


1. Roetenberg D, Slycke PJ, Veltink PH. Ambulatory position and orientation 

tracking fusing magnetic and inertial sensing. IEEE Trans Biomed Eng. 

2007;54(5):883-890. 

2. Roetenberg D, Luinge H, Baten CTM, Veltink PH. Compensation <strong>of</strong> 

Magnetic Disturbances Improves Inertial and Magnetic Sensing <strong>of</strong> Human 

Body Segment Orientation. 2005. Available at: http://doc.utwente.nl/53057/ 

[Accessed January 4, 2010]. 

3. Vries WD, Veeger H, Baten C, Helm FVD. Magnetic distortion in <strong>motion</strong> 

labs, implications for validating inertial magnetic sensors. Gait & Posture. 

2009;29(4):535-541. 

281

282

CHAPTER 7 

DATA VARIABILITY IN MOTION 

ANALYSIS 

ABSTRACT 

7.1 SEGMENTATION OF MOVEMENT 


ABSTRACT 

The aim <strong>of</strong> this chapter is to describe and discuss a method for the 

segmentation <strong>of</strong> movement developed and applied during the validation studies 

for the upper limb <strong>protocols</strong> <strong>based</strong> on stereophotogrammetry and inertial 

sensors. 

283

7.1 Segmentation <strong>of</strong> movement 

Extracted from: 

INTER-OPERATOR RELIABILITY AND PREDICTION 

BANDS OF A NOVEL PROTOCOL TO MEASURE THE 

COORDINATED MOVEMENTS OF SHOULDER-GIRDLE 

AND HUMERUS IN CLINICAL SETTINGS 

Gar<strong>of</strong>alo P, Cutti AG, Filippi MV, Cavazza S, Ferrari A, Cappello A, Davalli A 

Medical & Biological Engineering & Computing, 2009 May; 47(5):475-86 

Acronyms 

GD-H-R: girdle-humeral rhythm 

HAA: humerus ab-adduction movement 

HFE: humerus flexion-extension movement 

AA: humerus to thorax ab-adduction angle 

FE: humerus to thorax flexion-extension angle 

: FE mean value 

ED: shoulder-girdle to thorax elevation-depression angle 

PR: shoulder-girdle to thorax protraction-retraction angle 

: median <strong>of</strong> the values <strong>of</strong> ED correspondent to the onsets <strong>of</strong> the upward 

phases <strong>of</strong> HFE and HAA 

HT: high-threshold for FE(t) or AA(t) 

LT: low-threshold for FE(t) or AA(t) 

T: vector <strong>of</strong> time samples 

t EUP : time sample correspondent to the End <strong>of</strong> an Upward Phase <strong>of</strong> a movement 

t ODP : time sample correspondent to the Onset <strong>of</strong> a Downward Phase <strong>of</strong> a 

movement 

t EDP : time sample correspondent to the End <strong>of</strong> a Downward Phase <strong>of</strong> a 

movement 

t OUP : time sample correspondent to the Onset <strong>of</strong> an Upward Phase <strong>of</strong> a 

movement 

284


The output <strong>of</strong> the protocol described in [1], is a graphical representation <strong>of</strong> the 

GD-H-R during HFE and HAA. Specifically, the GD-H-R <strong>of</strong> HFE is described 

by 4 angle-angle plots, 2 for the upward phase <strong>of</strong> the movement (humerus 

moving cranially) and 2 for the downward phase (humerus moving caudally): 

ED vs FE & PR vs FE - upward phase, ED vs FE & PR vs FE – downward 

phase. Similarly, the GD-H-R <strong>of</strong> HAA is described by: ED vs AA & PR vs AA 

- upward phase, ED vs AA & PR vs AA – downward phase. 

To reach this representation <strong>of</strong> the GD-H-R, before being plotted FE, AA, ED 

and PR undergo two macro-steps, namely, (1) segmentation <strong>of</strong> the repetitions in 

the upward and downward phases, and (2) <strong>of</strong>fset removal from ED and PR 

patterns. 

The aim <strong>of</strong> this annex is to provide the details <strong>of</strong> the segmentation algorithm. 

To present a complete view <strong>of</strong> the two processing macro-steps, the <strong>of</strong>fset 

removal technique is also described briefly. 

2. ALGORITHM 

For the 5 consecutive repetitions <strong>of</strong> HFE, the segmentation consists in the split 

<strong>of</strong> the angle patterns <strong>of</strong> FE, ED and PR in the upward (humerus moving 

cranially) and downward (humerus moving caudally) phases <strong>of</strong> the movement. 

Similarly for the 5 consecutive repetitions <strong>of</strong> HAA, but for the angle patterns 

AA, ED and PR. 

The algorithm developed for the segmentation is original and it will be 

described hereinafter with explicit reference to HFE and the angles FE and ED. 

However, similar steps apply to FE and PR for HFE, as well as to AA and ED 

and to AA and PR for HAA. 

The segmentation algorithm consists <strong>of</strong> 3 steps (fig. 1). 

In step 1 (fig.1a) the FE angle versus time, FE(t), is considered and processed 

as follows (fig. 2): 

1a) the mean value <strong>of</strong> FE(t) is computed; all values <strong>of</strong> FE(t) above 

are defined as the ―high region‘s values‖ <strong>of</strong> FE(t) and the values below 

as the ―low region‘s values‖; 

1b) the lowest maximum <strong>of</strong> FE(t) in the high region is computed and it 

defines a high threshold for FE, named HT; 

1c) a low threshold (LT) is also defined equal to FE = 0°; 

285

1d) starting from the first sample, the samples <strong>of</strong> FE(t) are scanned until 

two samples are found such that FE(t 0 ) > LT and FE(t 0 -1) < LT. The 

sample <strong>of</strong> time t 0 is then classified as the onset <strong>of</strong> the first upward 

phase and it is stored in a time vector, T. 

1e) from t 0 to the end <strong>of</strong> the FE(t), all the frames t i with: 

 

FE(t i ) > HT and FE(t i -1) < HT are classified as the end <strong>of</strong> an 

upward phase (t EUP ); 

FE(t i ) < HT and FE(t i -1) > HT are classified as the onset <strong>of</strong> the 

downward phase (t ODP ); 

1f) the first sample in time t i : 

 

after a t ODP and such that FE(t i ) < LT and FE(t i -1) > LT is 

classified as the end <strong>of</strong> a downward phase (t EDP ); 

before a t EUP and such that FE(t i -1) < LT and FE(t i ) > LT is 

classified as the onset <strong>of</strong> a upward phase (t OUP ). 

1g) all time samples classified as t OUP , t EUP , t ODP and t EDP are stored in T in 

increasing order <strong>based</strong> on their values. By grouping the values in T by 

adjacent couples, FE(t) is divided in consecutive repetitions <strong>of</strong> upward 

and downward phases. 

In step 2, the last 4 upward and downward phases <strong>of</strong> FE(t) are selected and 

considered for the further steps (fig.1b). The time vector T is updated 

consequently and named T*. 

In step 3, the samples in T* are used to segment ED(t) in upward and 

downward phases (fig.1c). 

The algorithm to remove the <strong>of</strong>fset <strong>of</strong> ED (and PR) starts from the output <strong>of</strong> 

step 3. Firstly, the ED(t) values <strong>of</strong> the time samples classified as t OUP (onset <strong>of</strong> 

the upward phase) are considered and their median value, 

286 

, is computed. 

is then subtracted to ED(t). 

Once completed the segmentation and the <strong>of</strong>fset removal, corresponding 

upward and downward phases <strong>of</strong> FE and ED are plotted one versus the other to 

obtain the two angle-angle plots ED vs FE – upward phase, and ED vs FE – 

downward phase. 

The algorithm for segmentation and <strong>of</strong>fset removal is applied also for the 

segmentation <strong>of</strong> the HAA repetitions. However, for this movement the onset <strong>of</strong> 

the upward phase and the end <strong>of</strong> the downward phase are related to an LT <strong>of</strong> 

15° instead <strong>of</strong> 0°, in order to compensate for the difference between subjects in 

the AA angle value when the arms hang aside the body.

3. DISCUSSION 

The segmentation algorithm is original and represents a proposal to standardize 

the representation <strong>of</strong> angle-angle (coordination) plot in upper-extremity <strong>motion</strong> 

<strong>analysis</strong> study, which is currently lacking. In particular, the segmentation 

algorithm proposed presents three noticeable features: 

1) the algorithm discards all the oscillations <strong>of</strong> FE around 0° (or for AA around 

15°), occurring when the subject is relaxing his/her shoulder between a 

repetition <strong>of</strong> the movement and the next one; 

2) thanks to the use <strong>of</strong> a low and a subject-specific high threshold, the 

recognition <strong>of</strong> a repetition is not conditioned by local minima and maxima 

during the upward and the downward phases <strong>of</strong> the kinematics pattern (i.e. is 

not conditioned by the smoothness <strong>of</strong> the pattern); 

3) since the high threshold is parameterized for each subject on the minimum 

range <strong>of</strong> FE or AA common to all repetitions, the algorithm allows to take into 

account the maximum range <strong>of</strong> <strong>motion</strong> available common to all the repetitions, 

i.e. it allows to always take into account the maximum quantity <strong>of</strong> kinematic 

―information‖ available. 

287

Figure 1 Steps followed for the segmentation 

and <strong>of</strong>fset removal algorithm. Bold solid lines 

represent the upward phases, dot lines 

represent the downward phase. a) FE(t) is 

processed according to step 1; b) the last 4 

upward and downward phases <strong>of</strong> FE(t) are 

selected; c) time samples coming from the 

FE(t) segmentation are used to segment ED(t). 

is computed and 4) is subtracted to all 

ED(t); e) ED vs FE resulting from the 

segmentation procedure. The upward and 

downward angle-angle plots are here reported 

as a single plot. Dashed lines: downward 

phases; solid lines: upward phases. 

288

Figure 2 Step 1 <strong>of</strong> the segmentation procedure, applied to FE(t). Mean value, , <strong>of</strong> FE(t) is 

computed and low (LR) and high (HR) regions <strong>of</strong> the signal are defined. A high threshold (HT) is 

computed in HR and a low threshold (LT) is defined in LR. The segmentation algorithm allows to 

extract the time samples t OUP, t EUP, t ODP and t EDP, and therefore extract from FE(t) the upward and 

downward phases <strong>of</strong> the movement. 

289





2009;47(5):475-486. 

290

291

CHAPTER 8 

CONCLUSIONS 

The aim <strong>of</strong> this thesis was to describe the <strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> 

<strong>protocols</strong> for applications on upper and lower limb extremities, by using inertial 

sensors-<strong>based</strong> systems. Inertial sensors-<strong>based</strong> systems are relatively recent. 

Knowledge and <strong>development</strong> <strong>of</strong> methods and algorithms for the use <strong>of</strong> such 

systems for clinical purposes is therefore limited if compared with 

stereophotogrammetry. However, their advantages in terms <strong>of</strong> low cost, 

portability, small size, are a valid reason to follow this direction. When 

developing <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on inertial sensors, attention must 

be given to several aspects, like the accuracy <strong>of</strong> inertial sensors-<strong>based</strong> systems 

and their reliability. As discussed in Chapter 1, differently from the case <strong>of</strong> 

stereophotogrammetry, the knowledge about the human body and the specific 

application <strong>of</strong> the system, join to the ensemble <strong>of</strong> methods adopted for the 

estimation <strong>of</strong> the kinematic quantities, within the measurement system itself. 

Therefore, the need to develop specific algorithms/methods and s<strong>of</strong>tware for 

using these systems for specific applications, is as much important as the 

<strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on them. 

For this reason, the goal <strong>of</strong> the 3-years research project described in this thesis, 

was achieved first <strong>of</strong> all trying to correctly design the <strong>protocols</strong> <strong>based</strong> on 

inertial sensors, in terms <strong>of</strong> exploring and developing which features were 

suitable for the specific application <strong>of</strong> the <strong>protocols</strong>. The use <strong>of</strong> optoelectronic 

systems was necessary because they provided a gold standard and accurate 

measurement, which was used as a reference for the validation <strong>of</strong> the <strong>protocols</strong> 

<strong>based</strong> on inertial sensors and all the basic knowledge needful for the creation <strong>of</strong> 

a <strong>motion</strong> <strong>analysis</strong> protocol was the starting point for the generation <strong>of</strong> the same 

kind <strong>of</strong> knowledge in the case <strong>of</strong> inertial sensors. Among all the aspects, 

validation <strong>of</strong> <strong>protocols</strong> was a step required, being the intra-inter operator 

reliability an essential aspect to be evaluated for proposing the <strong>protocols</strong> as 

suitable for applications in clinical settings. From the clinical point <strong>of</strong> view, 

<strong>protocols</strong> <strong>based</strong> on stereophotogrammetry, validated during the doctorate, will 

be easily used for specific applications from rehabilitation centers in which 

optoelectronic systems are already used for research and/or clinical 

292

examinations. 

The <strong>protocols</strong> <strong>based</strong> on inertial sensors may be particularly helpful for 

rehabilitation centers in which the high cost <strong>of</strong> instrumentation or the limited 

working areas do not allow the use <strong>of</strong> stereophotogrammetry. Moreover, many 

applications requiring upper and lower limb <strong>motion</strong> <strong>analysis</strong> to be performed 

outside the laboratories, will benefit from these <strong>protocols</strong>, for example 

performing gait <strong>analysis</strong> along the corridors. Out <strong>of</strong> the buildings, the condition 

<strong>of</strong> steady-state walking or the behavior <strong>of</strong> the prosthetic devices when 

encountering slopes or obstacles during walking can also be assessed. 

The protocol presented here, <strong>based</strong> on stereophotogrammetry for the 

measurement <strong>of</strong> 3D shoulder kinematics in patients with shoulder pathology 

(Chapter 2), has been welcomed not only by INAIL Prostheses Centre (Vigorso 

di Budrio, Italy) but also by other research and rehabilitation centres like 

―Arcispedale S. Anna‖ (Ferrara, Italy). The s<strong>of</strong>tware created for the 

<strong>development</strong> and validation <strong>of</strong> the protocol was implemented by Aurion Srl 

(Milan, Italy) as ―Total3DUpperLimb‖ Vicon plugin that will be used together 

with the clinical tools available along with the Vicon system. 

The flexibility, simplicity and reliability <strong>of</strong> the <strong>motion</strong> <strong>analysis</strong> protocol <strong>based</strong> 

on MTx (Xsens Technologies B.V., The Netherlands) for the measurement <strong>of</strong> 

the scapulo-humeral rhythm (Chapter 5) was also welcomed by the Unit <strong>of</strong> 

Shoulder and Elbow Surgery, Cervesi Hospital (Cattolica, Italy) for 

applications on pre and post evaluation <strong>of</strong> surgery treatments. Another 

application, related to the monitoring <strong>of</strong> the shoulder on baseball pitchers 

during training seems also promising. 

The UDGEE (Unità di Riabilitazione delle Gravi Disabilità Infantili dell'Età 

Evolutiva) in Reggio Emilia (Italy), as well as the European centers VU 

Medisch Centrum (Amsterdam, The Netherlands), expressed interest on the 

lower limb protocol <strong>based</strong> on inertial sensors (Outwalk) (Chapter 4), as an 

ubiquitous system for 3D gait <strong>analysis</strong> for application on cerebral palsy 

children. 

INAIL Prostheses Centre stimulated and supported the <strong>development</strong> <strong>of</strong> 

additional methods for improving the accuracy <strong>of</strong> MTx in measuring the 3D 

kinematics for lower limb prostheses, with the results provided in this thesis. 

The application <strong>of</strong> inertial sensors on lower limb amputees presents conditions 

which are challenging for magnetometer-<strong>based</strong> systems, due to ferromagnetic 

material commonly adopted for the construction <strong>of</strong> idraulic components or 

motors. 

For transtibial amputees, a preliminary magnetic field mapping (Xsens manual) 

293

seems to be the solution, being the magnetic field produced by several screws 

homogenous. 

For transfemoral amputees, when these systems are supported by specific 

algorithms like KiC (Chapter 6), the system is valid and promising in terms <strong>of</strong> 

accuracy, providing a representative knee and ankle prostheses kinematics, 

although further steps are required to adopt the systems in clinical settings and 

adapt the use <strong>of</strong> the algorithm to specific sensor-to-segment calibration. 

INAIL Manager, the s<strong>of</strong>tware developed for the validation <strong>of</strong> upper and lower 

limb <strong>protocols</strong> <strong>based</strong> on MTx, is now a powerful and flexible tool for a fast and 

reliable application <strong>of</strong> the <strong>protocols</strong> in clinical settings and it can be a good 

starting point for future research at INAIL Prostheses Centre and future 

applications at INAIL local <strong>of</strong>fices in Italy. 

This thesis collects and summarizes various problems occurring when a 

researcher faces the creation <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong>, focusing on both 

theoretical and practical aspects implementation. The developing <strong>of</strong> the <strong>motion</strong> 

<strong>analysis</strong> <strong>protocols</strong> was partially far for creating <strong>protocols</strong> which can be adopted 

only by a specific instrumentation. Therefore, in theory, all the <strong>protocols</strong> 

presented here can be extended for their use with other instrumentations. 

Inertial sensors-<strong>based</strong> systems are suitable for the ambulatory measurement <strong>of</strong> 

3D kinematics, overcoming some <strong>of</strong> the limitations due to camera-<strong>based</strong> 

systems. However, this thesis demonstrated how the two worlds, 

optoelectronics and inertial sensors, are not "mutually exclusive". 

There is no apparent contrast between the two technologies and/or between 

different approaches in the creation <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong>. The upper 

limb <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> described in this thesis, for the 3D kinematics <strong>of</strong> 

the shoulder girdle using stereophotogrammetry and the measurement <strong>of</strong> the 

scapulo-humeral rhythm using inertial sensors, include completely different 

approaches from the methodological point <strong>of</strong> view (due to the limitations in the 

scapula tracking from stereophotogrammetry, and limitations in the 

representation <strong>of</strong> internal and external landmarks from inertial sensors-<strong>based</strong> 

systems). However, the creation <strong>of</strong> these <strong>protocols</strong> aim to the same goal, that is 

the availability <strong>of</strong> a tool for the measurement <strong>of</strong> the shoulder compensatory 

strategies. When using one with respect to the other, it can be decided 

depending on the clinical settings and the measurement requirements, for 

instance the kinds <strong>of</strong> activities the subject has to perform or whether the 

monitoring <strong>of</strong> the shoulder is included within a longitudinal study or not. 

The <strong>development</strong> <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> must be ―application-oriented‖, 

also considering that, as anticipated in the introduction, the knowledge about 

294

the subject/prosthesis, that is the object <strong>of</strong> the <strong>analysis</strong>, is necessary for the 

<strong>development</strong> and tuning <strong>of</strong> algorithms (in the case on inertial sensors); when 

high kinematic cross-talk occurs using anatomical approaches [1], the 

<strong>development</strong> <strong>of</strong> functional approaches must take into account the goal <strong>of</strong> the 

<strong>analysis</strong>. 

Therefore, rather than discussing about the possibility to adopt a ―general 

purpose‖ system for <strong>motion</strong> <strong>analysis</strong>, the author prefers to focus on other 

aspects, which are also further <strong>development</strong>s, like: 

- the need to augment the knowledge about how to use inertial sensors 

for <strong>motion</strong> <strong>analysis</strong> and how to improve their use in clinical settings, 

such as positioning over the body segments, s<strong>of</strong>t tissue artifact 

reduction, creation <strong>of</strong> sensor-to-segment calibration methods for 

specific body segments, and how to use inertial sensors together with 

EMG systems; 

- the need to augment the knowledge about how to optimize the use <strong>of</strong> 

stereophotogrammetry in terms <strong>of</strong> time required for the <strong>analysis</strong>, 

representation <strong>of</strong> the clinical outcome; 

- in general, the need to adapt <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> to be suitable in 

clinical settings, aligning the <strong>protocols</strong> outcome with the one 

commonly adopted by practitioners in order to diagnose pathologies or 

planning treatments. An example <strong>of</strong> this is provided by Cutti et al. [2], 

and Bold et al. [3], in which drug treatment, clinical evaluation scales 

and gait <strong>analysis</strong> are fused together into the rehabilitation process. 

Although the <strong>protocols</strong> described in this thesis already contain characteristics 

which align them with the clinical settings requirements, further <strong>development</strong>s 

are required to improve them. 

Not only the s<strong>of</strong>tware adopted is important, but also the data processing 

methods developed for analyzing results. This thesis described the need to 

segment the movement into different phases, proposing a method which takes 

into account other aspects like the <strong>of</strong>fset removing from 3D kinematics, which 

can be particularly critical when <strong>protocols</strong> are included into the decision 

making process [4,5]. 

Methods for the representation <strong>of</strong> lower limb amputees‘ kinematics (Chapter 3) 

can also be the starting point for realizing how the goals for the creation <strong>of</strong> 

295

methodology must be considered at the same level <strong>of</strong> methods implementation. 

An attempt to standardize these methods will not only augment the knowledge 

about their critical aspects, but will also produce comparable data among the 

researchers and allow the monitoring <strong>of</strong> the shoulder and lower limb 3D 

kinematics in time. 

In the author‘s opinion, this thesis and the <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on 

inertial sensors here described, are a demonstration <strong>of</strong> how a strict collaboration 

between the industry, the clinical centers, the research laboratories, can 

improve the knowledge, exchange know-how, with the common goal to 

develop new application-oriented systems. 

296

References 

1. Cutti AG, Gar<strong>of</strong>alo P, Davalli A, Cappello A. How accurate is the estimation 

<strong>of</strong> elbow kinematics using ISB recommended joint coordinate systems Gait & 

Posture. 2006;24:S36-S37. 

2. Cutti A, Gar<strong>of</strong>alo P, Filippi M. The evolution <strong>of</strong> compensation strategies in 

two patients with shoulder instability: a comparative study through quantitative 

<strong>motion</strong> <strong>analysis</strong>. Proc ISG2006 Chicago, USA. 2006. 

3. Boyd RN, Graham HK. Objective measurement <strong>of</strong> clinical findings in the 

use <strong>of</strong> botulinum toxin type A for the management <strong>of</strong> children with cerebral 

palsy. European Journal <strong>of</strong> Neurology. 1999;6(S4):s23-s35. 

4. Chau T, Young S, Redekop S. Managing variability in the summary and 

comparison <strong>of</strong> gait data. J Neuroeng Rehabil. 2005;2:22. 

5. Noonan KJ, Halliday S, Browne R, et al. Interobserver variability <strong>of</strong> gait 

<strong>analysis</strong> in patients with cerebral palsy. J Pediatr Orthop. 2003;23(3):279-287; 

discussion 288-291. 

297

298

CHAPTER 9 

PUBLICATIONS 

[1] Gar<strong>of</strong>alo P, Cutti AG, Raggi M, Davalli A: Knee kinematics 

measurement on above-knee amputees during gait in real-life environment 

using Inertial and Magnetic Measurement Units, accepted as oral presentation 

at ISPO 2010, 10-15 May 2010, Leipzig (Germany) 

[2] Cutti AG, Ferrari A, Gar<strong>of</strong>alo P, et al. 'Outwalk': a protocol for 

clinical gait <strong>analysis</strong> <strong>based</strong> on inertial and magnetic sensors. Med Biol Eng 

Comput. 2010;48(1):17-25. 

[3] Ferrari A, Cutti AG, Gar<strong>of</strong>alo P, et al. First in vivo assessment <strong>of</strong> 

"Outwalk": a novel protocol for clinical gait <strong>analysis</strong> <strong>based</strong> on inertial and 

magnetic sensors. Med Biol Eng Comput. 2010;48(1):1-15. 

[4] Gar<strong>of</strong>alo P, Cutti AG, Filippi MV, Cavazza S, Ferrari A, Cappello A, 

Davalli A: Inter-operator reliability and prediction bands <strong>of</strong> a novel protocol to 

measure the coordinated movements <strong>of</strong> shoulder-girdle and humerus in clinical 

settings, Medical & Biological Engineering & Computing, 2009 May; 

47(5):475-86 

[5] Gar<strong>of</strong>alo P, Raggi M, Ferrari A, Cutti AG, Davalli A: Development 

<strong>of</strong> a clinical s<strong>of</strong>tware to measure the 3D gait kinematics in every-day-life 

environment through the outwalk protocol, Proc. SIAMOC 2009, Gait & 

Posture, vol. 30, suppl. 1, p. 30 (October 2009) 

[6] Ferrari A, Cutti AG, Gar<strong>of</strong>alo P, Raggi M, Ferrari A: Outwalk: a new 

protocol to measure the 3D kinematics <strong>of</strong> gait in real-life environment using an 

inertial & magnetic measurement system, Proc. SIAMOC 2009, Gait & 

Posture, vol. 30, suppl. 1, pp. 52-53 (October 2009) 

[7] Fantozzi S, Gar<strong>of</strong>alo P, Cutti AG, Stagni R, Davalli A: Inverse 

dynamics Vs ground reaction force vector methods: application on lower limb 

299

amputees, Proc. SIAMOC 2009, Gait & Posture, vol. 30, suppl. 1, pp. 61-62 

(October 2009) 

[8] Cutti AG, Gar<strong>of</strong>alo P, Parel I, Fiumana G, Porcellini G: Intra- and 

inter-rater reliability <strong>of</strong> the scapulohumeral rhythm measured by a protocol 

<strong>based</strong> on inertial and magnetic sensors, Proc. SIAMOC 2009, Gait & Posture, 

vol. 30, suppl. 1, p. 17 (October 2009) 

[9] Cutti AG, Gar<strong>of</strong>alo P, Ferrari A, Raggi M, Cappello A: Development 

and test <strong>of</strong> a protocol <strong>based</strong> on an Inertial and Magnetic Measurement System 

to measure the 3D kinematics <strong>of</strong> gait in real-life environment, Proc. ESMAC 

2009, September 14-20, 2009, London, UK 

[10] Gar<strong>of</strong>alo P, Raggi M, Ferrari A, Cutti AG, Davalli A: Measure <strong>of</strong> the 

3D gait kinematics in real-life environments through the Outwalk protocol: 

Development <strong>of</strong> the end-user clinical s<strong>of</strong>tware. Gait & Posture. 2009;30:S132- 

S133 

[11] Cutti AG, Parel I, Gar<strong>of</strong>alo P, Fiumana G, Porcellini G: Intra- and 

inter-rater reliability <strong>of</strong> the scapulohumeral rhythm measured by a novel 

protocol <strong>based</strong> on inertial and magnetic sensors, Proc. ISB 2009, 5-9 July, 

2009, Cape Town, Sudafrica 

[12] Gar<strong>of</strong>alo P, Cutti AG, Parel I, Fiumana G, Porcellini G, Cappello A: 

Ambulatory measurement <strong>of</strong> the scapulohumeral rhythm: intra‐ and inter‐ rater 

reliability <strong>of</strong> a novel protocol <strong>based</strong> on inertial and magnetic sensors, Proc. 

Rehab Move, 4th International State‐<strong>of</strong>‐the‐art Congress, Rehabilitation: 

Mobility, Exercise & Sports, 7-9 April, 2009, Vrije Universiteit, Amsterdam 

[13] Fantozzi S, Gar<strong>of</strong>alo P, Cutti AG, Stagni R, Davalli A, Cappello A: 

Joint moments <strong>of</strong> the lower limb: inverse dynamics versus floor reaction force 

vector methods, Proc. ISPGR 2009, 21-25 June 2009, Bologna, Italy 

[14] Gar<strong>of</strong>alo P, Fantozzi S, Cutti AG, Tersi L, Ferrari A, Raggi M, Stagni 

R, Cappello A, Davalli A: Development <strong>of</strong> <strong>motion</strong> <strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on 

inertial sensors and fluoroscopy, Proc. 3D-MA 2008, 29 October 2008, 

Amsterdam, The Netherlands 

300

[15] Gar<strong>of</strong>alo P, Cutti AG, Filippi MV, Davalli A, Cappello A: Test-retest 

reliability <strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol for the assessment <strong>of</strong> shoulder-girdle 

compensatory movements, Proc. ISG 2008, 10-13 July 2008, Bologna, Italy 

[16] Gar<strong>of</strong>alo P, Cutti AG, Ludewig P, Phadke V, Cappello A: The 

relation between scapular and shoulder-girdle <strong>motion</strong> in subjects with and 

without shoulder impingement, Proc. ISG2008, 10-13 July, 2008, Bologna, 

Italy 

[17] Cutti AG, Gar<strong>of</strong>alo P, Ferrari A, Giovanardi A, Davalli A: Outdoor 

gait <strong>analysis</strong> using inertial and magnetic sensors: part 1 – protocol description, 

Proc. ICAMPAM 2008, 21-24 May, 2008, Rotterdam, The Netherlands 

[18] Ferrari A, Gar<strong>of</strong>alo P, Raggi M, Cutti AG, Cappello A: Outdoor gait 

<strong>analysis</strong> using inertial and magnetic sensors: part 2 – preliminary validation, 

Proc. ICAMPAM 2008, 21-24 May 2008, Rotterdam, The Netherlands 

[19] Cutti AG, Raggi M, Gar<strong>of</strong>alo P, Davalli A, Sacchetti A: The 

metabolic cost <strong>of</strong> two amputees walking outdoor with the ―power knee‖ 

prosthesis, Proc. ICAMPAM 2008, 21-24 May 2008, Rotterdam, The 

Netherlands 

[20] Fantozzi S, Giovanardi A, Camorani M, Cutti AG, Gar<strong>of</strong>alo P, Merni 

F: Kinematics <strong>analysis</strong> <strong>of</strong> a wheelchair tennis serve: a pilot study, Proc. ISG 

2008, 10-13 July 2008, Bologna, Italy 

[21] Gar<strong>of</strong>alo P, Cutti AG, Filippi MV, Cavazza S, Davalli A, Cappello A: 

Limitations <strong>of</strong> the constant scale in the assessment <strong>of</strong> shoulder compensatory 

strategies, Proc. SIAMOC 2007, Gait & Posture, vol. 28, suppl. 1, August 

2008, p. S29 

[22] Cutti AG, Gar<strong>of</strong>alo P, Filippi MV, Davalli A, Sacchetti R: Centro 

Protesi INAIL – Valutazione funzionale della spalla, submitted to Sphera 

Medical Journal, 2007 

301

[23] Cutti AG, Raggi M, Gar<strong>of</strong>alo P, Filippi MV, Davalli A, Sacchetti R: 

A <strong>motion</strong> <strong>analysis</strong> protocol for comparing active and reactive prosthetic knees, 

Proc. ISPO 2007, 29 July – 3 August 2007, Vancouver, CA 

[24] Cutti AG, Raggi M, Gar<strong>of</strong>alo P, Giovanardi A, Filippi MV, Davalli 

A: The effects <strong>of</strong> the ―Power Knee‖ prosthesis on amputees metabolic cost <strong>of</strong> 

walking and symmetry <strong>of</strong> gait – preliminary results, Proc. SIAMOC 2007, 

Gait & Posture, vol. 28, suppl. 1, August 2008, p. S38 

[25] Gar<strong>of</strong>alo P, Cutti AG, Davalli A, Cappello A: Inter-joint coordination 

patterns <strong>of</strong> able-bodied subjects for the <strong>analysis</strong> <strong>of</strong> compensatory strategies in 

patients with shoulder impairments, Proc. ISB 2007, Journal <strong>of</strong> 

Biomechanics, 2007, vol. 40, suppl. 2, p. S106 

[26] Cutti AG, Giovanardi A, Gar<strong>of</strong>alo P, Rocchi L, Davalli A: Motion 

<strong>analysis</strong> <strong>of</strong> the upper-limb <strong>based</strong> on inertial sensors: part 2 - preliminary 

validation <strong>of</strong> a novel protocol, Proc. ISB 2007, Journal <strong>of</strong> Biomechanics, 

2007, vol. 40, suppl. 2, p. S544 

[27] Cutti AG, Giovanardi A, Gar<strong>of</strong>alo P, Rocchi L, Davalli A: Motion 

<strong>analysis</strong> <strong>of</strong> the upper-limb <strong>based</strong> on inertial sensors: part 3 – assessment <strong>of</strong> 

instrumental accuracy for a new protocol, Proc. ISB 2007, Journal <strong>of</strong> 

Biomechanics, 2007, vol. 40, suppl. 2, p. S545 

[28] Cutti AG, Gar<strong>of</strong>alo P, Janssens K, Davalli A, Sacchetti R: 

Biomechanical <strong>analysis</strong> <strong>of</strong> an upper limb amputee and his innovative 

myoelectric prosthesis: a case study concerning the Ottobock ―Dynamic arm‖, 

Orthopaedie Technik (Quarterly), 2007, Issue 1, 6-15 

[29] Cutti AG, Gar<strong>of</strong>alo P, Davalli A, Cappello A: How accurate is the 

estimation <strong>of</strong> elbow kinematics using ISB recommended joint coordinate 

systems, Proc. ―First Joint ESMAC–GCMAS Meeting (JEGM06)‖, Gait & 

Posture, vol. 24, suppl. 2, December 2006, pp. S36-S37 

[30] Gar<strong>of</strong>alo P, Cutti AG, Filippi MV, Davalli A, Cappello A: Motion 

<strong>analysis</strong> in the assessment <strong>of</strong> rotator cuff tears impairment: a case study, Gait 

& Posture, vol. 24, suppl. 1, November 2006, p. S39 

302

[31] Cutti AG, Gar<strong>of</strong>alo P, Davalli A: Control <strong>of</strong> the myoelectric 

prosthesis and <strong>of</strong> the shoulder in above-elbow amputees: a case study, Proc. 

ISG 2006, 9-10 October, 2006, Chicago, US 

[32] Cutti AG, Gar<strong>of</strong>alo P, Filippi MV, Cavazza S, Davalli A, Cappello A: 

The evolution <strong>of</strong> compensation strategies in two patients with shoulder 

instability: a comparative study through quantitative <strong>motion</strong> <strong>analysis</strong>, Proc. 

ISG2006, 9-10 Ottobre 2006, Chicago, US 

[33] Cutti AG, Gar<strong>of</strong>alo P, Janssens K, Davalli A, Sacchetti R: A 

biomechanical <strong>analysis</strong> <strong>of</strong> upper-limb prostheses: performance assessment <strong>of</strong> 

the new Ottobock DYNAMIC ARM, Proc. ORTHOPÄDIE + REHA- 

TECHNIK, International Trade Show and World Congress for 

Prosthetics, Orthotics and Rehabilitation Technology, 10-13 May 2006, 

Leipzig, Germany 

[34] Schipper L, Roetenberg D, Gar<strong>of</strong>alo P, Cutti A, Luinge HJ: A method 

for gait <strong>analysis</strong> using inertial sensors, accepted as poster presentation at JEGM 

2010, Leipzig (Germany) 

[35] Ulrich MJH, van Tuijl EAB, Cutti AG, Gar<strong>of</strong>alo P, Veeger D: 

Ambulatory measurement <strong>of</strong> the scapulothoracic <strong>motion</strong>: accuracy <strong>of</strong> a protocol 

<strong>based</strong> on inertial and magnetic sensors, submitted to ISG 2010, 25-27 July 

2010, Minnesota (USA) 

[36] Cutti AG, Raggi M, Gar<strong>of</strong>alo P, Bottoni G, Amoresano A : 3D gait 

kinematic <strong>of</strong> transtibial amputees walking in every-day-life environments: 

reliability study <strong>of</strong> a protocol <strong>based</strong> on inertial & magnetic sensors, accepted as 

poster presentation at ISPO 2010, 10-15 May 2010, Leipzig (Germany) 

[37] Roetenberg D, Schipper L, Gar<strong>of</strong>alo P, Cutti AG, Luinge H: Joint 

angles and segment length estimation using inertial sensors, submitted to 

3DMA 2010, July 14-16 2010, San Francisco (US) 

303

304

About the author 

Personal information 

Place <strong>of</strong> birth: San Donà di Piave (Venice) 

Date <strong>of</strong> birth: 01/11/1979 

Education 

 

 

 

2007-2009: XXII Ph.D. Course in Bioengineering at DEIS 

(Department <strong>of</strong> Electronics, Computer Science and Systems), 

University <strong>of</strong> Bologna; Ph.D. thesis: ―Development <strong>of</strong> <strong>motion</strong> 

<strong>analysis</strong> <strong>protocols</strong> <strong>based</strong> on inertial sensors‖ 

December 2005: Laurea (5-years degree) on Electronic Engineering 

(Biomedical specialization) at University <strong>of</strong> Bologna; thesis: ―A 

Matlab toolbox for the upper limb functional evaluation through 

<strong>motion</strong> <strong>analysis</strong>‖, at INAIL Prostheses Centre, Vigorso di Budrio 

(BO) 

April- May 2005: Course on “Tools for the creation <strong>of</strong> an 

innovative enterprise”, 11 a edition, SINFORM, Bologna, Italy 

Pr<strong>of</strong>essional experiences 

 

 

 

 

January 2010 – now: Full time contract as Technical Product 

Manager – Movement Science area, at Xsens Technologies B.V. 

(The Netherlands) (http://www.xsens.com) 

October 1, 2008 – October 31, 2009: internship at Xsens 

Technologies B. V. (The Netherlands), ―Validation <strong>of</strong> Xsens inertial 

sensors technologies for the evaluation <strong>of</strong> orthopedic prostheses‖ 

January 1, 2008 – December 31, 2008: Sole administrator <strong>of</strong> 

TecnoPro Italia Cooperative Society, Bologna, Italy 

May 1, 2006 – April 30, 2007: Research contract at DEIS, University 

<strong>of</strong> Bologna, for the "Identification and validation <strong>of</strong> indices <strong>of</strong> motor 

305

performance <strong>of</strong> subjects with upper limb pathologies and <strong>development</strong> 

<strong>of</strong> a <strong>motion</strong> <strong>analysis</strong> protocol for transfemoral amputees" 

 

 

March 1, 2006 – April 30, 2006: Contract <strong>of</strong> collaboration at DEIS, 

University <strong>of</strong> Bologna, Italy 

December 2004 – December 2005: Research contract through the 

global grant ―Spinner‖, D.3 and D.4, for the creation <strong>of</strong> an 

underwater positioning system 

Complementary education 

 

Course on DOE (Design <strong>of</strong> Experiment) from Eng. G. Olmi, ―Scuola 

di Dottorato in Ingegneria Industriale‖, November - December 2007 

Course on Vicon Polygon, Aurion S.r.l. (MI), February 29, 2008 

 

 

 

 

―Workshop Kwon3D”, ISB 2007, July 4, 2007, Taipei, Taiwan 

X SIAMOC Conference (Societa' italiana di Analisi del Movimento 

in Clinica), October 1-3, 2009, Alghero, Italy 

ESMAC 2009 (European Society <strong>of</strong> Movement Analysis in Adults 

and Children), September 14-19, 2009, London, UK 

ISPGR 2009 (International Society for Posture and Gait Research), 

June 21-25, 2009, Bologna, Italy 

Rehab Move, 4th International State‐<strong>of</strong>‐the‐art Congress, 

Rehabilitation: Mobility, Exercise & Sports, April 7-9, 2009, Vrije 

Universiteit, Amsterdam, The Netherlands 

 

 

Jubileum Congres <strong>of</strong> the Roessingh 60 years about the 

Rehabilitation Technologies, December 12, 2008, Enschede, The 

Netherlands 

―10th Meeting <strong>of</strong> the technical group on '3D Analysis <strong>of</strong> Human 

Movement' <strong>of</strong> the International Society <strong>of</strong> Biomechanics", October 

29-31 2008, Amsterdam, The Netherlands 

―2 nd Summer School on Advanced technologies for neuro-motor 

assessment and rehabilitation‖, July 13-19, 2008, Monte San Pietro 

(Bologna), Italy 

 

International Shoulder Group (ISG 2008) – Shoulder Biomechanics, 

July 10-12, 2008, Bologna, Italy 

306

―International conference on Ambulatory Monitoring <strong>of</strong> Physical 

Activity and Movement (ICAMPAM 2008)‖, May 21-24 2008, 

Rotterdam, The Netherlands 

Innovat&Match 2008, Research To Business, June 5-6 2008, 

Bologna 

Innovact 2008, ―European Forum for Innovative growth companies‖ - 

March 18-19, 2008, Reims, France 

 

 

 

 

 

Seminar from Dr. Matteo Cioni on ―The Parkinson desease‖, Faculty 

<strong>of</strong> Engineering, University <strong>of</strong> Bologna, December 2007, Bologna, 

Italy 

Seminar on ―New realization <strong>of</strong> the control unit for the INAIL 

upper limb functional prosthesis‖, from Eng. Fabio Cacciari at 

INAIL Prostheses Centre, Vigorso di Budrio (BO), November 2007 

Seminar on ―Active should prosthesis with 2 degrees <strong>of</strong> freedom: 

from the design to the experimental characterization‖, from Eng. 

Marco Chiossi, at INAIL Prostheses Centre, Vigorso di Budrio 

(BO), October 2007 

VIII SIAMOC Conference, “Analisi del movimento in clinica‖, 

October 24-27, 2007, Cuneo, Italy 

XXVI Scuola Annuale di Bioingegneria, ―Genomica e proteomica 

computazionale‖, September 24-28, 2007, Bressanone (BZ), Italy 

International Society <strong>of</strong> Biomechanics, ISB 2007, July 1-5, 2007, 

Taipei, Taiwan 

 

 

 

 

 

 

Seminar on ―Shoulder dynamic study‖, from Eng. Saulo Martelli, at 

―Laboratorio di Tecnologia Medica‖, ―Istituti Ortopedici Rizzoli‖, 

May 2007, Bologna, Italy 

―Corso EMG di superficie”, at Istituti Ortopedici Rizzoli, November 

30 – December 2, 2006, Bologna, Italy 

―SIAMOC 2006: L'analisi del movimento nel processo decisionale 

clinico”, October 18-21, 2006, Empoli, Italy 

―First joint Meeting <strong>of</strong> ESMAC & GCMAS – JEGM06 ”, Vrije 

Universiteit, September 27-30, 2006, Amsterdam, The Netherlands 

First Summer School on ―Advanced technologies for the neuromotor 

evaluation and rehabilitation‖, DEIS, University <strong>of</strong> Bologna, June 19- 

24, 2006, Bologna, Italy 

Exposanità 2006, “Mostra Internazionale al servizio della sanità e 

della salute‖, May 2006, Bologna, Italy 

307

ORTHOPAEDIE + REHA-TECHNIK, May 21-24, 2006, Leipzig, 

Germany. 

Course on Linux at Faculty <strong>of</strong> Enegineering, University <strong>of</strong> Bologna, 

Bologna, Italy 

―2nd European School <strong>of</strong> Neuroengineering Massimo Grattarola‖, 

June 2004, Genova, Italy 

Course on “First aid techniques”, Associazione Seirs, Exposanità 

2004, May 2004, Bologna, Italy 

Exposanità 2004, Mostra Internazionale al servizio della sanità e della 

salute, May 2004, Bologna, Italy 

Course on ―Skills balancing‖, ARSTUD and FONDAZIONE 

ALDINI VALERIANI, in collaboration with PROFINGEST, 

SER.IN.AR. and CE.TRANS, 2004, Bologna, Italy 

XXII Scuola Annuale “Bioingegneria della Postura e del 

Movimento“, September 22-25, 2003, Bressanone (BZ), Italy 

Bionova 2003, June 4-6, 2003, Padova, Italy 

Seminars on ―Neurosensory coding <strong>of</strong> Movement‖ at DEIS, 

University <strong>of</strong> Bologna, March 25, 2003, Bologna, Italy 

III SIAMOC Conference and Tutorial ―Human <strong>motion</strong> <strong>analysis</strong> 

through stereophotogrammetry‖ at ―Istituti Ortopedici Rizzoli‖, 

October 13-15, 2002, Bologna, Italy 

Bionova 2001, November 28 – December 1, 2001, Padova, Italy 

Extra activities 

 

 

 

 

PADI “Open Water Diver” patent for underwater activities 

Applicant to an italian and european patent about an underwater 

positioning systems for SCUBA diving applications 

Participation to a student organization at Faculty <strong>of</strong> Engineering, 

Bologna, Italy, with the creation <strong>of</strong> a Linux User Group, G.L.I.B. 

(http://linux.ing.unibo.it) 

Delegate <strong>of</strong> students at the Faculty <strong>of</strong> Engineering, Bologna, Italy 

Trainer <strong>of</strong> young baseball players 

308

309

Ringraziamenti 

Il completamento di questa tesi ha suscitato in me particolare soddisfazione, in 

quanto prova della mia crescita personale e pr<strong>of</strong>essionale. Debolezze, paure, 

ansie, che un tempo animavano la mia quotidianità da studente, sono state 

decimate dall‘unione di attività, eventi, persone, comunioni durante il mio 

dottorato. Questa tesi non e‘ che la testimonianza di un lungo percorso 

cominciato con tante aspettative nei confronti di Bologna e di me stesso. 

Posso con sicurezza dire che l‘esperienza del dottorato sia qualcosa di molto 

complicato ma altrettanto indimenticabile. 

Quando i miei amici mi chiedevano come fosse fare il dottorato, spesso 

raccontavo che fare il dottorato e‘ un po‘ come fare immersioni in mare: se non 

lo provi, non puoi mai capire come sia, nei pro e nei contro. Seppur possa 

essere difficile da comprendere, così come immergendosi in acqua ci si sente 

sospesi, liberi di muoversi in tutte le direzioni ma al tempo stesso con delle 

limitazioni e delle regole da seguire per giungere al punto di arrivo, ecco che il 

dottorato si presenta allo stesso modo: come dottorando puoi ricercare qualsiasi 

cosa possa esserti utile o affascinante, una ricerca potenzialmente senza limiti, 

operando delle scelte che hanno delle conseguenze dirette o indirette sul 

proprio lavoro, ma al tempo stesso ricercare tenendo sempre presente quale sia 

lo scopo finale del proprio lavoro. 

Attraverso questa similitudine è forse possibile comprendere come il dottorato 

sia in grado di soddisfare il proprio desiderio di libertà, formando però al tempo 

stesso una propria personalità che abbia al suo interno anche doti come la 

rigorosità dei propri metodi, l‘approccio critico alle cose della vita, la ricerca, 

molto ambiziosa, di una soluzione semplice ad ogni tipo di problema. 

Tuttavia, e‘ stato per me quasi sorprendente accorgermi di quanto la mia 

personalità si sia formata anche in virtù delle mie aspirazioni personali, delle 

cose in cui credo, delle mie idee che andavano al di là di ciò che studiavo. In 

questo, il dottorato ha sicuramente avuto la funzione di catalizzatore. 

A differenza delle immersioni in mare, a fare il dottorato non ci si sente mai 

soli. Ci si sente facente parte di una squadra, che lotta per uno scopo comune, 

fino alla fine e, nel mio caso, anche dopo. Per questo motivo riporto questi 

ringraziamenti in duplice copia, in italiano ed inglese. Questa tesi e‘ infatti il 

310

isultato dell‘unione delle forze di così tante persone messe insieme, che queste 

pagine difficilmente potrebbero riuscire a ringraziarle veramente. E‘ in 

quest‘ottica che ringrazio l‘Italia ma anche i Paesi Bassi. L‘Italia per aver 

disposto tutte le basi necessarie affinché le mie capacità potessero soddisfare i 

requisiti del dottorato, e per aver creato la connessione attualmente esistente 

con il paese dei tulipani. 

Ancora prima dell‘inizio del mio dottorato, in Italia sono stato immediatamente 

accolto e responsabilizzato al DEIS dal pr<strong>of</strong>. Angelo Cappello ed al Centro 

Protesi INAIL dall‘ Ing. Andrea Giovanni Cutti, la cui tesi di dottorato insieme 

alla mia, sono il risultato di un enorme lavoro svolto insieme in questi anni con 

dedizione e passione. Con l‘Ing. Cutti ed il suo ineguagliabile incipit ―Si 

potrebbe‖ la ricerca non si e‘ mai fermata ai ―ma‖, ai ―forse‖ ed inoltre non e‘ 

mai stata fine a se stessa, ma aveva sempre come traguardo quello di ―creare 

qualcosa che potesse servire‖. Tutto questo grazie anche al supporto di tutto il 

Laboratorio di Analisi del Movimento del Centro. Nella lista, senza pretendere 

di essere esaustivo, troviamo Michele Raggi, che con la sua ironia, tra un 

labelling ed un altro, ha saputo strappare una risata anche nei momenti più 

difficili della mia ricerca; Alberto Ferrari, detto anche ―Ferris‖, con molta più 

aerodinamicità dell‘omonima casa automobilistica, e con il quale sono sempre 

riuscito ad avere un dialogo pr<strong>of</strong>icuo; Emanuele Gruppioni, maestro circuitale e 

di vita, che con la sua saggezza mi ha spesso illuminato; Ilaria Parel, Maria 

Vittoria Filippi, Andrea Giovanardi e tanti altri. A loro devo sicuramente 

moltissimo, in quanto ―compagni di ventura‖ per molti anni. Senza dimenticare 

tutti gli studenti belgi e quelli olandesi che ci hanno affiancato. 

Non dimentico inoltre i Paesi Bassi per avermi accolto splendidamente non 

soltanto dal punto di vista pr<strong>of</strong>essionale ma anche umano. L‘unione di questi 

due paesi in questa ―avventura‖ e‘ risultata non soltanto pr<strong>of</strong>icua ma ha anche 

evidenziato come l‘Italia sia un paese dalle enormi potenzialita‘ che, se ben 

canalizzate, possono produrre un risultato notevole per la comunita‘ scientifica, 

e per la vita in generale. I Paesi Bassi, dal canto loro, accolgono calorosamente 

le idee e le aspettative altrui. In questo senso, Italia e Paesi Bassi formano 

sicuramente un‘accoppiata vincente. Ringrazio quindi l‘azienda Xsens 

Technologies B.V. di Enschede, a partire dai fondatori Casper Peeters e Per 

Slycke; piu‘ in particolare l‘area ricerca capitanata da Henk Luinge e tutto il 

suo magnifico gruppo – Linda, Kiman, Marijke, Job, Frans, Patrick - e tanti 

altri: mi hanno supportato e sopportato durante un lungo percorso che ha 

permesso di creare un valido ponte tra Italia ed i Paesi Bassi. 

311

Un particolare ringraziamento va a Silvia ed alla piccola Eleonora, per essermi 

stati vicini nell‘ultima parte della mia lunga ricerca scientifica e non solo. A 

loro devo parecchio, anche perche‘ mi hanno aiutato a porre ordine nel mondo 

che mi circondava. 

Un caloroso ringraziamento va anche a mio cugino Nino, con il quale avrei 

molto volentieri condiviso questo momento, ma sono sicuro che proprio adesso 

stia leggendo queste righe e sorridendo come ha sempre saputo fare. 

Infine, dedico interamente quest‘ultima parte alla mia famiglia piu‘ stretta, i 

miei cari papa‘ e mamma e mia sorella Francesca, per poi arrivare ai miei cari 

nonni. Tutti loro hanno sopportato la mia distanza da casa. I miei genitori 

hanno creduto nelle mie capacita‘ fin dall‘inizio e mi hanno supportato per 

l‘intero arco della mia esperienza lontano dalla Terra dei Limoni (sebbene 

alcuni dicano delle arance). Con estrema sicurezza affermo che senza di loro 

avrei comunque scritto queste righe, ma non sarei mai arrivato a scriverle in 

questa tesi. Certamente esserci riusciti conferma quanto siano sempre stati degli 

splendidi genitori. Sono riusciti a non porre freno al mio desiderio di conoscere 

il mondo, seppur al tempo stesso avvertendomi di cio‘ che avrei incontrato 

lungo il mio cammino. Ed ora posso dire che questa via di mezzo ha dato i suoi 

frutti. 

Un grazie caloroso a tutti i miei amici sparsi per il mondo...a loro devo tanti bei 

ricordi e la certezza che in qualsiasi posto mi dovessi spostare troverei 

qualcuno con il quale scambiare i propri mondi, le proprie conoscenze ed i 

propri sogni. 

―Esistono due gruppi di persone, quelle che sognano, e quelle che non lo 

ammettono‖ 

312

313

Acknowledgments 

I was particularly pleased when I was preparing and <strong>of</strong> course finishing this 

thesis, as pro<strong>of</strong> <strong>of</strong> my personal and pr<strong>of</strong>essional growth. Weaknesses, fears, 

anxieties, once surrounding my everyday life as a student, have been defeated 

by activities, events and people during my PhD. This thesis is definitely the 

evidence <strong>of</strong> a long journey which started with great expectations from Bologna 

and from myself. I can certainly assert that the experience <strong>of</strong> the Ph.D. is 

something very complicated but equally memorable. 

When my friends asked me how it could be to be a Ph.D. candidate, I <strong>of</strong>ten said 

that being a Ph.D. candidate is like to make SCUBA diving: if you do not try it, 

you can never understand how it is, pros and cons included. Although it might 

be difficult to understand, just like diving into the water you feel suspended, 

free to move in all directions but at the same time with the limitations and rules 

you must follow to reach the point <strong>of</strong> arrival, well…here as a Ph.D. candidate 

you can do the same, you can look for anything that can help you or that is 

intriguing, a research potentially with no limits, making choices that have direct 

or indirect consequences on your work, but at the same time focusing on the 

ultimate goal <strong>of</strong> your entire work. ―Change your objectives, but focus on the 

goal‖, we might say. Through this comparison it is perhaps possible to 

understand how a Ph.D. candidate can satisfy his/her desire for freedom, but at 

the same time forming his/her own personality which even contents gifts such 

as conformity to rules in the methods, a critical approach during life, the 

research, very ambitious, <strong>of</strong> the fastest problem solving in every situation. 

However, and it was surprising for me to realize how my personality was 

transformed thanks to my personal aspirations, my ideas were sometimes 

beyond what I was studying, although, surprisingly, a lot <strong>of</strong> times they were 

very close. In all <strong>of</strong> this, the Ph.D. has certainly served as a catalyst. Unlike 

SCUBA diving, the Ph.D. will not make you feel alone. You feel part <strong>of</strong> a 

team, fighting for a common purpose, to the end and, in my case, even after. 

For this reason these acknowledgments are here in Italian and English 

languages. This thesis is de facto the result <strong>of</strong> forces coming from so many 

people together, that these lines could hardly be enough to thank them. 

With this spirit I thank Italy but also the Netherlands. Italy for having provided 

all necessary for my abilities to fulfill the requirements <strong>of</strong> the Ph.D., and for 

creating the currently existing connection with the Country <strong>of</strong> Tulips. Even 

314

efore the beginning <strong>of</strong> my Ph.D., in Italy, I was immediately welcomed and 

my accountability was augmented by pr<strong>of</strong>. Angelo Cappello from DEIS who 

introduced me Eng Andrea Giovanni Cutti now at INAIL Prostheses Centre, 

whose Ph.D. thesis together with mine are the result <strong>of</strong> a very hard work 

together over the years with dedication and passion for biomechanics and 

clinics. With Cutti and his peerless incipit "We definitely could…" our research 

did never stop at "but", or "maybe" and a goal was always there, that is to 

create ―something useful‖. All this happened thanks to the support <strong>of</strong> all the 

Laboratory <strong>of</strong> Motion Analysis at INAIL. Among the elements <strong>of</strong> the list, 

without claiming to be exhaustive, we can find Michele Raggi, who with his 

irony, between one marker labeling and the other one, he was able to make me 

laugh even in the most difficult periods <strong>of</strong> my research; Alberto Ferrari, also 

known as "Ferris", with much higher dynamics than the homonymous car 

brand, and with whom I have always managed to have a fruitful dialogue; 

Emanuele Gruppioni, the master <strong>of</strong> circuits and life, who with his wisdom I 

have <strong>of</strong>ten meet ―Satori‖; Ilaria Parel, Maria Vittoria Filippi, Andrea 

Giovanardi and many others. I am very obliged to all <strong>of</strong> them, as "companions 

<strong>of</strong> fortune" for many years. I do not forget all the Dutch and Belgian students 

who have joined our research and routine. I do not forget the Netherlands for 

magnificently welcoming me, not only from a pr<strong>of</strong>essional but also human 

point <strong>of</strong> view. 

The union <strong>of</strong> these two countries in this adventure was not only pr<strong>of</strong>itable but it 

also turned out that Italy is a country <strong>of</strong> huge potential which, if properly 

channeled, can produce a remarkable outcome for all the scientific community 

and life in general. The same for the Netherlands, because they welcome ideas, 

expectations and they are the fountain <strong>of</strong> pragmatism. In this sense, Italy and 

the Netherlands are certainly winning pairing. 

I thank Xsens Technologies B.V., first <strong>of</strong> all the founders Per Slycke and 

Casper Peeters, then in particular the research area led by Henk Luinge and all 

his wonderful group - Linda, Kiman, Daniel, Marijke, Makoto, Joroen, Job, 

Frans, Patrick - and many others: I was supported and endured by them during 

a long journey which allowed to create an effective bridge between Italy and 

the Netherlands. 

Special thanks to Sylvia and the young Eleonora, for being close to the last part 

<strong>of</strong> my long scientific research and even more than this. I really thank you, 

because you helped me to make order into my life. 

315

Special thanks to my cousin Nino, whom I would have gladly shared this 

moment, but I am sure that right now he is reading these lines and smiling as he 

has always been doing…and he will do. 

Finally, this is a dedication to my family, my dear papa and mama and my 

sister Francesca, then dear great grandparents. All <strong>of</strong> them have suffered my 

distance from home, but my parents believed in my abilities since the beginning 

and supported me throughout all my experience away from the Land <strong>of</strong> Lemons 

(although they say Oranges). For sure, without them I would have written these 

lines, but I would have never come to write this thesis. To be here confirms 

how wonderful my parents are. They managed to not slow down my desire to 

know the world, although they have always been warning me about what I 

could find along my way. And now I can say that this trade<strong>of</strong>f was fruitful. 

Special thanks to all my friends around the world... I have so many memories 

<strong>of</strong> you and I am sure that any place I move to I would find someone to 

exchange our worlds, our knowledge and..our dreams. 

―There are two groups <strong>of</strong> people: people dreaming, and people not admitting it‖ 

316

development of motion analysis protocols based on inertial ... - Xsens

Create successful ePaper yourself

Delete template?

Save as template?