Clinical evaluation of monitor unit software and the application of ...

Dose calculation
Clinical evaluation of monitor unit software
and the application of action levels
Dietmar Georg a, * ,1 , Tufve Nyholm b,1 ,Jörgen Olofsson b , Flemming Kjær-Kristoffersen c ,
Bruno Schnekenburger d , Peter Winkler e ,Ha˚kan Nyström c ,
Anders Ahnesjö b,f,g , Mikael Karlsson b
a Department of Radiotherapy, Medical University Vienna/AKH Vienna, Austria, b Department of Radiation
Sciences, Radiation Physics, Umea˚ University, Sweden, c Department of Radiation Oncology, Copenhagen University Hospital, Denmark,
d Department of Radiation Oncology, University Hospital Basel, Switzerland, e Department of Therapeutic Radiology and Oncology, Medical
University Graz, Austria, f Department of Oncology, Radiology and Clinical Immunology, Uppsala University, Sweden,
g Nucletron Scandinavia, Uppsala, Sweden
Abstract
Purpose: The aim of this study was the clinical evaluation of an independent dose and monitor unit verification (MUV)
software which is based on sophisticated semi-analytical modelling. The software was developed within the framework
of an ESTRO project. Finally, consistent handling of dose calculation deviations applying individual action levels is
discussed.
Materials and methods: A Matlab-based software (‘‘MUV’’) was distributed to five well-established treatment centres
in Europe (Vienna, Graz, Basel, Copenhagen, and Umea˚) and evaluated as a quality assurance (QA) tool in clinical
routine. Results were acquired for 226 individual treatment plans including a total of 815 radiation fields. About 150
beam verification measurements were performed for a portion of the individual treatment plans, mainly with time
variable fluence patterns. The deviations between dose calculations performed with a treatment planning system (TPS)
and the MUV software were scored with respect to treatment area, treatment technique, geometrical depth, radiological
depth, etc.
Results: In general good agreement was found between calculations performed with the different TPSs and MUV, with
a mean deviation per field of 0.2 ± 3.5% (1 SD) and mean deviations of 0.2 ± 2.2% for composite treatment plans. For
pelvic treatments less than 10% of all fields showed deviations larger than 3%. In general, when using the radiological
depth for verification calculations the results and the spread in the results improved significantly, especially for headand-neck
and for thorax treatments. For IMRT head-and-neck beams, mean deviations between MUV and the local TPS
were 1.0 ± 7.3% for dynamic, and 1.3 ± 3.2% for step-and-shoot IMRT delivery. For dynamic IMRT beams in the pelvis
good agreement was obtained between MUV and the local TPS (mean: 1.6 ± 1.5%). Treatment site and treatment
technique dependent action levels between ±3% and ±5% seem to be clinically realistic if a radiological depth correction
is performed, even for dynamic wedges and IMRT.
Conclusion: The software MUV is well suited for patient specific treatment plan QA applications and can handle all
currently available treatment techniques that can be applied with standard linear accelerators. The highly sophisticated
dose calculation model implemented in MUV allows investigation of systematic TPS deviations by performing calculations
in homogeneous conditions.
c 2007 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology 85 (2007) 306–315.
Keywords: Dose calculation accuracy; Action level; Independent dose calculation
Technical developments in radiotherapy, such as multileaf
collimators (MLC) and three dimensional (3D) treatment
planning systems, have enabled complex treatment techniques
and have given us the opportunity to escalate doses
to targets without increasing dose burdens to surrounding
1 These authors contributed equally to the work.
Radiotherapy and Oncology 85 (2007) 306–315
www.thegreenjournal.com
healthy tissues. In the complex world of modern radiation
oncology, dose calculation performed with a computerized
treatment planning system (TPS) represents one of the most
essential steps in the treatment chain, as the only realistic
technique to estimate dose delivery in situ. Although limitations
of dose calculation algorithms are present in most
commercial treatment planning systems, systematic reports
0167-8140/$ - see front matter c 2007 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.radonc.2007.04.035

of these limitations are limited and practical guidelines for
QA of TPS have become available only recently. From previous
ESTRO projects on quality assurance (QA) aspects in
radiotherapy that include the treatment planning system
(e.g. ESTRO-EQUAL or QUASIMODO) it can be concluded that
there are uncertainties related to the dose calculation models
[4,7,9]. Another example is the study of Venselaar and
Welleweerd [21], where it is shown that especially asymmetric
wedged beams are not properly modelled in some
commercial treatment planning systems.
On the other hand, it is generally accepted that deviations
larger than 5% between delivered and prescribed doses
may influence the clinical outcome of the treatment [2,14].
For specific tumours and situations, such as in clinical trials,
even higher demands might be required in terms of dosimetric
accuracy. Because errors and large uncertainties in dose
calculations reduce the quality of a treatment, independent
dose or MU calculations have been recommended and also
used for a long time as a routine quality assurance (QA) procedure
when verifying individual treatment plans
[3,11,15,16]. During the last years commercial products
providing independent dose calculations that can handle
all advanced treatment techniques have become available.
However, to our knowledge no reports have been published
that describes their accuracy or other aspects of their clinical
application.
Verification calculations are traditionally performed by
applying empirical algorithms in a manual calculation procedure,
or utilizing software based on fairly simple dose calculation
algorithms. With the advent of advanced treatment
techniques that are based on the availability of multileaf
collimators, asymmetric jaws and dynamic or virtual
wedges, empirical dose calculation procedures become
much more complex and their accuracy is often limited by
the simplicity of the model itself [6]. To achieve high accuracy
with an independent dose calculation tool also for the
most complex treatment techniques, more sophisticated
models are required [12,13,22]. On the other hand, only
models that allow detailed investigations of the dosimetric
accuracy, and consequently enable more strict action levels
(AL), will be effective in catching systematic uncertainties
of the TPS. Moreover, misinterpretations of TPS results,
and the occurrence of errors associated with the limited
accuracy of empirical verification models can be greatly
reduced.
As a general requirement, an ideal verification dose calculation
model should be as independent as possible of the
TPS, and should be based on physical effects which are
accurately described and tuned by an independent set of
algorithm input data [10]. Additionally, the verification
software implementation should provide high accuracy
uncorrelated to the radiotherapy equipment in use, i.e. its
accuracy should neither depend on the type of linear accelerator
nor the photon beam energy for which it has been
commissioned. To be able to verify multiple beams in an
efficient way, it should manage to import treatment plan
data (e.g. MLC settings) directly from the TPS or the record
and verify (R&V) system.
Ideally, a full QA-procedure should include verification of
all parameters including the patient anatomy and positioning.
In the procedure suggested here the patient geometry
D. Georg et al. / Radiotherapy and Oncology 85 (2007) 306–315 307
is verified separately on an everyday basis. The MU verification
tool will thus be used only prior to the treatment start,
provided that the integrity of the database can be secured
by other methods during the duration of the treatment
period.
The overall intention in modern radiation therapy is to
keep the dose delivered to the patient as close as possible
to the prescribed dose, i.e. to fulfil a tight dosimetric tolerance
level. Such a dosimetric tolerance level should be
based on clinical considerations related to radiobiological
parameters. In order to secure that a treatment falls within
this dosimetric tolerance level at a given probability an action
level must be derived. This action level will then also
be dependent on the accuracy of the verification tool.
Hence, for a given dosimetric tolerance level and a known
uncertainty in the verification, the action level that should
be applied for that individual treatment plan can be estimated.
Correctly applied, this concept will always result
in an action level that is narrower than the dosimetric tolerance
level. An MU software should therefore be designed to
give a high degree of accuracy, including an estimation of
the overall uncertainty in the dose calculation.
The aim of the present work was to test and evaluate a
fluence-based independent dose calculation (or MU) algorithm
(for point dose calculations) against calculations from
various commercial 3D treatment planning systems under
clinical conditions. For that purpose clinically accepted
and approved treatment plans were considered. For a subgroup,
deviations were further analysed by measurements.
The clinical applications covered open beams, wedged
beams with both physical and dynamic wedges, and stepand-shoot
as well as dynamic IMRT. The underlying model
for independent calculations was implemented in such a
way that only a few input data are needed for commissioning
and tuning; thus the workload for basic algorithm input
data acquisition and commissioning is reduced to a minimum.
Furthermore, the risk for commissioning errors is largely
reduced with a minimum number of input data. By
analysing the resulting deviations between dose calculations
performed with a TPS and MUV we have tried to assess the
application of action levels for independent dose checks
for advanced conformal radiotherapy.
Methods and materials
Independent dose calculation software MUV
The semi-analytical model implemented in the software
MUV, which stands for monitor unit verification, was based
on a two step procedure: (i) calculation of the energy fluence
per monitor unit exiting the treatment head, and (ii)
calculation of the dose deposition in the phantom resulting
from the incident energy fluence. The energy fluence part
in the dose calculation takes into account the direct energy
fluence, scattered radiation from an extra-focal source and
scattered fluence from secondary collimators, as well as
backscattered radiation to the monitor chamber [19,20].
The dose deposition was based on a pencil beam model, with
a radial parameterisation [1,17,18]. For dose calculations,
collimator transmission and the effect of rounded leaf ends

308 Clinical evaluation of fluence based MU software
are taken into account. These geometries were modelled
equally if they occurred in IMRT, or in asymmetric fields.
The model and the software were developed with the
intention of creating a dose per MU verification tool that requires
a minimum of commissioned input data. For tuning
the fluence model for open beams, only 10 measured output
factors in air are recommended and the pencil beam model
is solely based on the beam quality index TPR20,10 [17,19].
For physical wedges, the wedge angle needs to be specified
by the user. In addition, the position of the wedge in the
treatment head (i.e. distance from the focal source) needs
to be given and two wedge factors need to be measured at
10 cm depth for two different field sizes, e.g. for
10 · 10 cm 2 and 20 · 20 cm 2 . From all this information a lateral
wedge modulation matrix, a lateral wedge scatter matrix,
and a lateral shift of beam quality are derived and used
in the calculations.
Another novel feature of the dose verification tool MUV is
its uncertainty estimation. The uncertainty prediction was
based on thorough testing of the individual components of
the model against measurements under different treatment
conditions for a variety of existing treatment machines. For
example, for open beams the calculation model for the total
energy fluence provided results within 1% (2 SD) [19] and
for irregular MLC shaped beams the total output factor
(OF total) had an average absolute deviation of 0.4% and a
maximum deviation of 1.7% [20]. This was lower than the
deviations found in a comparable investigation of a commercial
treatment planning system [8]. The pencil beam
model in MUV includes effects of lateral beam quality variations,
which is not the case for most commercial TPSs
according to the knowledge of the authors [18]. A more detailed
description of the experimental benchmarking tests,
which included head scatter calculations on and off axis in
open fields, dose calculations on axis in irregular MLC
shaped fields at different depths, and the influence of offaxis
beam softening (including pencil beam corrections)
can be found elsewhere e.g. [17–20]. The actual uncertainty
estimation implemented in MUV was based on collimator/MLC
settings and treatment depth. Uncertainty
estimations are desirable from any dose calculation module
because they can be helpful when defining action levels for
clinical QA routine. However, the clinical implementation of
action levels depending on a semi-analytical uncertainty
analysis and subsequent automatic warning of the user requires
further investigations and is beyond the scope of
the present study.
A DICOM interface allows extracting and importing of all
dosimetric (e.g. energy, MU) and geometric (e.g. leaf and
jaw positions) treatment data for an entire treatment plan
instantaneously and automatically, which is a pre-requisite
for an efficient QA process for more complex treatment
techniques, such as conformal therapy utilizing an MLC or
IMRT. The final result of the dose calculation in the MUV
software is the dose to water in a pre-defined point of interest
in a homogeneous phantom, for a composite treatment
plan as well as for each individual beam. The depth of the
dose specification (normalisation) point in a treatment plan
that is automatically exported by a TPS is the geometric
depth. If a simple inhomogeneity correction is desirable
because of the actual treatment situation, the calculation
depth can be modified, e.g. by using some sort of radiological
path length correction with respect to the dose specification
point. In cases where it was obvious that depth
corrections were needed, the radiological depth was used
for calculations with MUV. The radiological depth (or equivalent
path length – EPL) was calculated manually by measuring
distances and dimensions of bulk densities directly
on the final treatment plan, or could be taken from the
TPS if verified. If taken from the TPS, spot checks were
made and the agreement between manual and computerized
calculation was generally within 1–2 mm, which did
not influence the final results. The point of interest in the
patient can be modified in all directions in MUV by the user
(anterior–posterior, lateral and longitudinal) so that finally
multiple points can be verified if needed.
Radiotherapy equipment
The independent monitor unit verification software
‘‘MUV’’ was distributed to five well-established radiotherapy
departments across Europe (Medical University Vienna, Medical
University Graz, University Hospital Basel, University
Hospital Copenhagen, and University Umea˚). Algorithm input
data were measured for dedicated accelerators in various
photon beams. For beam configuration in MUV, the treatment
head configuration (i.e. distances and thickness of field defining
elements, position of the physical wedge, etc.) could be
selected from a pre-defined linac database which covers all
state-of-the-art accelerator brands. The database was established
taking vendor information into account and by measuring
certain distances directly on the respective treatment
machine. The centres and the respective equipment (treatment
planning systems including applied algorithms and linear
accelerators) used are listed in Table 1. All linear
accelerators were equipped with an MLC and represent the
most common types used clinically today.
For verification of calculations in homogeneous conditions
measurements were performed with ionisation chambers
in water phantoms or solid verification phantoms. At
the Medical University Vienna, measurements were carried
out in a Solid Water TM phantom with a calibrated ionisation
chamber (Nuclear Enterprises, type 2611A, Volume
0.3 cm 3 ) connected to a NE electrometer (Nuclear Enterprises,
type 2620). In Basel a Farmer type chamber (volume
0.6 cm 3 PTW Freiburg, Germany) connected to PTW Unidos
electrometer was used in a scanning water phantom, while
in Graz a small volume ionisation chamber (PTW, type
31002, 0.125 cm 3 ) was used for verification measurements
in water. The Copenhagen centre used the same type of
chamber but placed in a Solid Water TM phantom.
Clinical testing of MUV
In each centre, MUV was evaluated as a QA tool in clinical
routine, i.e. MUV calculations were compared with
calculations performed by the local TPS. As typically done
for such independent verification calculations, either a
semi-infinite water phantom was assumed or the calculations
were performed in a dedicated solid verification
phantom. Note that for the involved TPS, only full 3D
treatment plans based on CT information were considered
in this study.

Table 1
Overview of institutions and the respective radiotherapy equipment for clinical testing and evaluation of the independent dose verification
software ‘MUV’
Radiotherapy institution TPS and applied algorithm Linacs Energies (MV)
University Umea˚ Helax TMS (V6.1) pencil beam or collapsed cone a
Siemens Primus 6/18
Medical University Vienna Helax TMS (V6.1) pencil beam or collapsed cone a
ELEKTA Precise 6/10/25
GE SAT 43 6/10/25
Medical University Graz Pinnacle (V6.2b) superposition Varian Clinac 2300 CD 6/18
University Basel XiO (V4.2) convolution/superposition ELEKTA Precise 6/18
University Copenhagen Eclipse (V7.3.10) pencil beam Varian Clinac 2300 EX 6/18
a
...applied for lung treatments only.
Results were acquired for 226 individual treatment plans
including a total of 815 radiation fields. Table 2 gives an
overview of the considered cases as a function of treatment
technique and treatment area. All different treatment techniques,
ranging from regular and irregular open fields to
wedged fields using a physical or dynamic wedge, and dynamic
as well as step-and-shoot IMRT could be handled with
MUV and were included in this study.
The accuracy of MUV against measurements, performed
in a homogeneous water phantom, has been demonstrated
during the design and pre-clinical testing phase [17–20].
For example, for irregular MLC shaped fields a standard
deviation of 0.47% was obtained between calculated and
measured output factors (in water or air) for about 300 test
Table 2
Summary of treatment fields, treatment techniques, treatment
sites, etc., for clinical testing of the independent dose verification
software ‘MUV’
Treatment area # of plans # of fields Techniques
Pelvis 98 364 Open 273
Physical wedge 46
Dynamic wedge 38
Step-shoot IMRT 0
Dynamic IMRT 7
Thorax 46 155 Open 95
Physical wedge 17
Dynamic wedge 36
Step-shoot IMRT 7
Dynamic IMRT 0
Head-and-neck 71 271 Open 75
Physical wedge 48
Dynamic wedge 22
Step-shoot IMRT 63
Dynamic IMRT 63
Other 11 25 Open 3
Physical wedge 20
Dynamic wedge 2
Step-shoot IMRT 0
Dynamic IMRT 0
Total 226 815 Open 446
Physical wedge 131
Dynamic wedge 98
Step-shoot IMRT 70
Dynamic IMRT 70
D. Georg et al. / Radiotherapy and Oncology 85 (2007) 306–315 309
cases, where 5, 10, and 20 cm depth were considered together
with four different MLC designs. Maximum deviations
did not exceed 1.7%, even for the most irregular field shapes
[20]. Therefore, the main focus of this study was the clinical
application of the independent MUV software. However,
verification measurements were also performed for a small
number of individual treatment plans encompassing in total
150 beams. Of these more than 100 beams with segmentally
modulated fluence patterns (dynamic wedges and IMRT)
were specifically tested as they were not considered in detail
in previous tests.
Finally, all observed deviations between MUV and the different
TPSs were analysed in order to assess the clinical
application of realistic action levels (AL). For that purpose
current results and previously performed experimental
benchmarking tests of the algorithm implemented in MUV
were combined [17–20].
Results
In general good overall agreement was found between
calculations performed with the different TPS and MUV calculations,
with a mean deviation per field of 0.2 ± 3.5%
(1 SD), and mean deviations of 0.2 ± 2.2% for a composite
treatment. For a more detailed analysis all treatments
and/or fields were binned with respect to treatment area,
treatment technique, and the use of geometrical depth or
radiological depth, respectively.
Treatment site specific deviations
Fig. 1a and b present the fraction of beams with deviations
between the local TPS and MUV that are exceeding a
certain limit, separated for treatment site (pelvis, thorax,
head-and-neck, other) and whether any depth corrections
were performed. While for pelvic treatments less than 10%
of all fields showed deviations larger than 3%, irrespective
of radiological depth corrections, this fraction was almost
40% for thorax fields and 30% for head-and-neck fields.
When using the radiological depth, the fraction of beams
with deviations larger than 3% could be reduced to about
10% for head-and-neck and to 30% for thorax treatments.
For the latter application an agreement between TPS and
MUV calculations for about 90% of all fields could be
achieved only at the 4.5% deviation level (when applying
the radiological depth).

310 Clinical evaluation of fluence based MU software
a
Exception fraction
b
Exception fraction
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
When analysing only those fields where information on
both depths was available, the following mean values and
standard deviations were obtained when using geometric
or radiological depth: 1.1 ± 5.8% versus 0.8 ± 1.7% for pelvic
fields (total 157), 3.7 ± 6.6% versus 1.5 ± 2.1% for thorax
fields (total 65) and 0.5 ± 7.5% versus 0.8 ± 1.4% for
head-and-neck fields (total 83), respectively. Fig. 2 illustrates
the improvements for radiological depth corrections
and shows median and quartiles for these applications.
Treatment technique specific deviations
Table 3 summarizes the observed deviations between
calculations performed with MUV and with the local TPS
as a function of treatment technique and treatment site.
Columns ‘‘all data’’ refer to calculations where either the
ALL DATA
Pelvic
Thorax
Head & Neck
Other
ALL sites
0 1 2 3 4 5 6 7 8 9 10
Deviation [%] (MUV-TPS)/MUV
ONLY RAD DEPTH DATA
Pelvic
Thora
Head & Neck
ALL sites
0 1 2 3 4 5 6 7 8 9 10
Deviation [%] (MUV-TPS)/MUV
Fig. 1. Fraction of beams with dose deviations that were exceeding a certain deviation limit, for the major treatment areas. (a) Using input
data based on either geometric or radiological depth, (b) using input data based on radiological depth only.
geometric or the radiological depth was used while columns
denoted as ‘‘RL’’ refer to calculations where solely information
on the radiological depth was applied in independent
calculations. Verification calculations performed for
wedged beams with physical wedges were obviously done
using radiological path length correction, because mean values
and standard deviation are very similar in columns ‘‘all
data’’ and ‘‘RL’’, irrespective of treatment site. This was
not the case for open beams and dynamic wedges; these
techniques show much better agreement when using RL
(for comparison see columns for thorax fields and headand-neck
fields). In general, slightly larger deviations were
obtained for wedged beams compared to open beams.
In all five centres contributing to this study the most frequent
IMRT application was head-and-neck, where each 63

DEVIATION [%]
6%
4%
2%
0%
-2%
-4%
-6%
-8%
-10%
Pelvic fields
(157)
fields were checked for step-and-shoot IMRT delivery and
dynamic MLC IMRT delivery. Mean deviations (incl. standard
deviations) between MUV and the local TPS were
1.0 ± 7.3% for dynamic IMRT delivery and 1.3 ± 3.2% for
step-and-shoot IMRT delivery, respectively. Step-and-shoot
IMRT results in the thorax were again biased by tissue inhomogeneity;
no dynamic IMRT was performed in the thoracic
region. For dynamic IMRT cases in the pelvis good agreement
was obtained between MUV and the local TPS (mean:
1.6 ± 1.5%).
In order to verify IMRT fields against ionisation chamber
measurements and to compare the achievable accuracy
for IM fields with the one for open beams, a few treatment
plans were recalculated in a homogeneous phantom with
both MUV and the TPS. In other words, recalculated IMRT
verification plans were considered in this part. Table 4 summarizes
the results for open static beams and IMRT cases,
both on an individual field basis and for composite treatment
plans. These results confirm the high accuracy of the
independent dose/MU calculation software MUV in homogeneous
conditions, i.e. when comparing ionisation chamber
measurements with dose calculations in identical conditions
D. Georg et al. / Radiotherapy and Oncology 85 (2007) 306–315 311
Thoracic fields
(65)
Head & neck
fields (83)
Fig. 2. Illustration of the improvements for radiological depth corrections at the field level. Median and quartiles for pelvic fields (total 157),
thoracic fields (total 65) and head-and-neck fields (total 83) when using geometric or radiological depth for independent dose calculations
(filled circles...with radiological path length correction, filled diamonds...all data).
Table 3
Summary of deviations between MUV and the local treatment planning system, as a function of conformal treatment technique and
treatment site (RL... radiological depth)
Pelvic fields Thorax fields Head-and-neck fields
All data RL data All data RL data All data RL data
Open beams 0.5 ± 1.3% (273) 0.6 ± 1.3% (107) 2.6 ± 7.6% (95) 1.6 ± 2.2% (51) 0.5 ± 2.1% (75) 1.0 ± 1.1% (45)
Physical wedges 0.9 ± 2.3% (46) 1.3 ± 2.6% (31) 1.1 ± 1.0% (17) 1.2 ± 0.9% (8) 0.6 ± 1.5% (48) 1.1 ± 1.5% (24)
Dynamic wedges 0.2 ± 2.6% (38) 1.3 ± 1.5% (19) 3.2 ± 7.8% (36) 0.6 ± 2.5% (6) 0.0 ± 2.0% (22) 0.3 ± 1.4% (14)
The numbers in brackets indicate the total number of fields per category
in a water phantom or a solid verification phantom without
homogeneities. There are, however, additional uncertainties
in both TPS calculations and calculations performed
with MUV, due to the extent and accuracy in modelling
rounded leaf ends, tongue-and-groove effects, leaf transmission
or the distribution of the direct source (X-ray target).
When compared to open beams, larger mean
deviations and standard deviations for individual IMRT fields
(see Table 4a) are also influenced by the larger overall
experimental uncertainty for ionisation chamber measurements
in IMRT treatments [24]. Large relative deviations
are also influenced by the beam contribution to the overall
treatment plan and for composite treatment plans excellent
agreement was found between ionisation chamber measurements
and calculations performed with MUV, see Table 4b.
Clinical action levels
Fig. 3a and b show treatment site dependent frequency
distributions of deviations between MUV and TPS calculations.
The dashed lines indicate a deviation level of ±3% or
±5%. For all treatment sites there were systematic deviations
which are biased by dose calculation uncertainties

312 Clinical evaluation of fluence based MU software
Table 4
Comparison of dose calculations and ionisation chamber (IC) measurements for individual treatment fields performed under identical
conditions in a water phantom or a solid water equivalent verification phantom
Treatment technique
(a) Individual fields
Number of fields (MUV IC)/IC (mean ± 1 SD) (%) (TPS IC)/IC (mean ± 1 SD) (%)
Open irregular fields 47 0.7 ± 0.6 0.3 ± 1.0
Dynamic IMRT 56 0.7 ± 4.1 1.0 ± 5.8
Step-and-shoot IMRT 48 0.6 ± 2.5 1.1 ± 3.3
Number of plans (MUV IC)/IC (mean ± 1 SD) (%) (TPS IC)/IC (mean ± 1 SD) (%)
(b) Composite treatment plan
Open irregular fields 12 0.4 ± 0.6 0.4 ± 1.1
Dynamic IMRT 9 0.1 ± 1.1 1.5 ± 3.0
Step-and-shoot IMRT 7 0.3 ± 1.0 1.7 ± 0.9
and the fact that dose calculations in a homogeneous flat
water equivalent medium were compared with dose calculations
that were based on CT information. Best agreement
between TPS and independent dose calculations was observed
for pelvic fields. When using radiological depth information
the majority (98%) of all deviations was within ±5%,
for pelvic and head-and-neck fields even within ±3% ( 92%
of all fields).
From the results of ionisation chamber measurements
presented in Table 4a and from previously published experimental
verification of MUV in homogeneous conditions,
treatment technique dependent confidence intervals were
defined as follows. For open beams the confidence interval
in a homogeneous medium was even as small as ±2%. However,
for the implementation of clinical action levels additional
uncertainties need to be considered that account
for different dose calculation conditions for the TPS and
independent verification calculations, i.e. effects related
to patient anatomy and inhomogeneities. These additional
tolerances were estimated from the results presented in
Figs. 1–3 and Table 3. When applying radiological depth
corrections a clinical action level of ±3% (open) and ±5%
(wedged and IMRT), respectively, seems to be suitable for
pelvic and head-and-neck treatments. For thorax treatments,
even with radiological depth correction, a site specific
additional uncertainty of 3% seems to be necessary
for all treatment techniques. If no radiological depth correction
is performed site dependent additional uncertainties
of 2% are recommended for head-and-neck treatments
and 5% for thorax treatments.
Discussion
Dose calculations performed with a treatment planning
system (TPS) can be verified with different approaches.
Although quality assurance (QA) of TPS has been recognized
as an important topic for many years, practical guidelines or
recommendations for performance verification and periodic
QA of TPS have become available only recently [10,15,16].
The detection of systematic errors in TPS calculations requires
extensive or ‘‘stress’’ testing, which is difficult to
perform in practice because no specific recommendation
is available addressing the number of tests, test conditions,
etc. On the other hand, if systematic uncertainties under
specific conditions remain overlooked, the quality of a
radiotherapy treatment is limited and no corrective actions
can be set. An intelligent design to carry out such stress
tests in an efficient way for a particular TPS requires a priori
knowledge of the dose calculation algorithm. For those reasons
stress tests are very often based on experimental techniques.
In that aspect dedicated dose calculation software,
which is designed to achieve as high dose calculation accuracy
in homogeneous conditions as possible, can be very
helpful in order to keep the workload at an acceptable level.
From the results obtained within the present study
and from previously published articles from the same
authors [17–20], we conclude that MUV can be considered
as such a tool.
Besides systematic TPS tests and QA during the commissioning
phase it is also recommended to use a second independent
MU calculation to verify the MU data produced by
the TPS as part of patient specific QA. For many years
empirical factor based formalisms or in-vivo dosimetry have
been in use for independent dose verification e.g. [3,5].
These procedures were governed by fairly simple treatment
techniques with rather uncomplicated planning procedures.
In the light of advanced treatment techniques based on MLC
technology and complex planning procedures the general
properties of independent dose calculation need to be
reconsidered. Demands on accuracy could be shifted towards
the standards of today’s planning dose calculations
or even higher; single point dose verification in a semi-infinite
slab approximation could be replaced by 2D (or 3D) calculations
taking into account the patients’ anatomy. In
other words, independent dose calculation algorithms can
be designed to provide a high level of accuracy, even when
patient anatomy is included. Depending on how well the
independent dose calculation is integrated into the clinical
workflow long calculation times may not constitute an actual
problem. For example, QA-procedures for dose calculation
can be designed to run automatically, i.e. without
demanding any manual operations, by executing verification
calculations on all treatment plans prior to the treatment
start from a ‘‘QA robot’’, which is for example linked to
the database of the record and verify (R&V) system. Treatment
plans exceeding a certain deviation could be marked
in the database and a notification is sent to clinical physicists.
This is also an attractive concept to fulfil QA and pa-

a
b
Frequency
Frequency
120
100
80
60
40
20
70
60
50
40
30
20
10
0
0
-11%
-10%
-11%
-10%
-9%
-8%
-7%
-6%
-5%
-4%
-3%
-2%
tient safety regulations because all treatment plans would
be subjected to automated independent dose calculations.
Ideally, such an automated procedure is followed by a retrospective
analysis that continuously scans through the
database looking for systematic sources of dose calculation
error/uncertainties. This may consequently be an additional
QA strategy where small but systematic error/uncertainties,
either in the planning calculation or in the independent dose
calculation, can be found and possibly also explained and
fixed. However, for such an automated QA-procedure it is
still necessary to check the treatment plan by an experi-
D. Georg et al. / Radiotherapy and Oncology 85 (2007) 306–315 313
All data
Deviation (MUV-TPS)/MUV [%]
Only radiological depth data
-9%
-8%
-7%
-6%
-5%
-4%
-3%
-2%
-1%
0%
1%
2%
3%
Deviation (MUV-TPS)/MUV [%]
thorax
HN
pelvic
-1%
0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
RD thorax
RD HN
RD pelvis
enced professional in order to avoid hot spots, dose prescription
and plan normalisation errors. Independent dose
calculations based on export files from TPS do not check
correct data file transfer from the TPS to the treatment unit
and the actual performance of the treatment unit, which
are other sources of error in the radiotherapy chain. However,
these aspects could be included by measuring leaf settings
and MU measured with an EPID [23]. Moreover, such
information could be used as well as input information for
an independent dose calculation. Finally, there is also a tangible
risk for errors to be introduced by the user(s) during
10%
4%
5%
6%
7%
8%
9%
10%
Fig. 3. Treatment site dependent frequency distributions of deviations between dose calculations performed with MUV and the local TPS. (a)
Using input data based on either geometric or radiological depth, (b) using input data based on radiological depth only.

314 Clinical evaluation of fluence based MU software
the planning and preparation procedure [25], often due to
simple human mistakes in combination with inappropriate
clinical routines.
As long as patient anatomy is not included in verification
calculations the accuracy of independent dose checks is influenced
by treatment site specific factors. This is the most
striking limitation of independent dose calculation procedures
that are employed in clinics today. For some treatment
areas, such as thorax and head-and-neck, accurate results
cannot be achieved in simple calculation conditions, i.e. a
semi-infinite homogeneous phantom. With radiological depth
corrections, for head-and-neck treatments almost as good results
as for pelvic treatment could be achieved (see Fig. 1a
and b). Interestingly, the impact of radiological depth correction
was limited for pelvic treatments. This might be biased
by the class solutions, i.e. beam geometry, applied in the centres
participating in the study. Although for thorax fields the
agreement between TPS and independent MU calculations
was largely improved when using radiological depth information,
the overall agreement was inferior compared to the
other treatment areas. This can be explained by the overestimated
scatter when performing calculations in a water equivalent
homogeneous medium and comparing them with
calculations accounting for differences in scatter conditions
from inhomogeneous media.
When defining action levels for any QA process the uncertainties
of procedures or calculations, which are used to derive
a certain value, need to be considered. A proper
selection of action level is a compromise between probabilities
of dose deviations outside the clinical tolerance level
and workload generated by false alarms. To ensure narrow
tolerance levels with an acceptable workload it is important
to use a verification tool capable of high accuracy. Further,
when the accuracy of the verification tool is dependent on
irradiation parameters inherent uncertainty estimation will
help optimising the QA-procedure.
Unfortunately, uncertainties are not accounted for in an
open and sincere manner, even though they are inevitable
when aiming to model complex processes such as dose deposition
in a patient. A well-known fact concerning computer
calculations is that in practice some sort of ‘‘priority-compromise’’
needs to be found between speed and accuracy. In that
aspect the importance of presenting also flaws and weaknesses
of a dose calculations model by the companies that
market TPSs must be emphasized. Such issues are often discussed
in a scientific context, but probably not as frequently
in the clinical routine work. Because the accuracy of independent
dose verification depends on input data (e.g. geometric
versus radiological depth), treatment area (e.g. missing tissue),
and treatment specific beam data (e.g. beam weight
to overall plan, weight of wedge) clinical AL should be defined
treatment site and treatment technique specific.
Conclusion
The software MUV is well suited for patient specific treatment
plan QA applications and can handle all currently
available treatment techniques that can be applied with
standard linear accelerators, such as regular and irregular
open beams, physical and dynamic wedges, step-and-shoot
and dynamic IMRT. The highly accurate dose calculation
model implemented in MUV allows investigating systematic
TPS deviations by performing calculations in homogeneous
conditions.
Information on the uncertainty from any dose calculation
algorithm would both enhance and simplify the process
of setting clinical action levels that identify cases
where the discrepancies between planning and independent
dose calculations need follow up. However, further
research is needed in that aspect. Based on the current
findings treatment site and treatment technique dependent
action levels between 3% and 5% seem to be clinically
realistic if a radiological depth correction is
performed.
Acknowledgements
This work was partly supported by ESTRO and performed within
the framework of the ESQUIRE project and partly supported by the
Cancer Research foundation in Northern Sweden.
* Corresponding author. Dietmar Georg, Division of Medical
Radiation Physics, Department of Radiotherapy, Medical University
Vienna/AKH Vienna, Währinger Gürtel 18-20, A-1090 Vienna,
Austria. E-mail address: Dietmar.Georg@akhwien.at
Received 13 September 2006; received in revised form 9 March
2007; accepted 24 April 2007; Available online 27 September 2007
References
[1] Ahnesjö A, Saxner M, Trepp A. A pencil beam model for photon
dose calculation. Med Phys 1992;19:263–73.
[2] Brahme A. Dosimetric precision requirements in radiation
therapy. Acta Radiol Oncol 1984;23:379–91.
[3] Dutreix A, Svensson H, Bjärngard BE, Bridier A, Mijnheer A,
Shaw J. Monitor unit calculation for high energy photon
beams, Physics for clinical radiotherapy. Booklet No. 3,
ESTRO; 1997.
[4] Ferreira IH, Dutreix A, Bridier A, Chavaudra J, Svensson H. The
ESTRO-QUALity assurance network (EQUAL). Radiother Oncol
2000;55:273–84.
[5] Fiorino C, Corletto D, Mangili P, Broggi S, Bonini A, Cattaneo
GM, et al. Quality assurance by systematic in vivo dosimetry:
results on a large cohort of patients. Radiother Oncol
2000;56:85–95.
[6] Georg D, Olofsson J, Künzler T, Karlsson M. On empirical
methods to determine scatter factors for irregular MLC shaped
beams. Med Phys 2004;31:2222–9.
[7] Gillis S, De Wagter C, Bohsung J, Perrin B, Williams P, Mijnheer
BJ. An inter-centre quality assurance network for IMRT
verification: results of the ESTRO QUASIMODO project. Radiother
Oncol 2005;76:340–53.
[8] Hansson H, Bjork P, Knöös T, Nilsson P. Verification of a pencil
beam based treatment planning system: output factors for
open photon beams shaped with MLC or blocks. Phys Med Biol
1999;44:N201–7.
[9] Heeren G. The bright but ephemeral life of a rainbow: a
chronicle of seventeen years of intensive ESTRO-EU cooperation.
Radiother Oncol 2005;75:253–7.
[10] IAEA Technical Reports Series No. 430: Commissioning and
quality assurance of computerized planning systems for radiation
treatment of cancer. International atomic energy
agency, IAEA; 2005.

[11] Knöös T, Johnsson SA, Ceberg CP, Tomaszewicz A, Nilsson
P. Independent checking of the delivered dose for highenergy
X-rays using a hand-held PC. Radiother Oncol
2001;58:201–8.
[12] Linthout N, Verellen D, Van Acker S, Storme G. A simple
theoretical verification of monitor unit calculation for intensity
modulated beams using dynamic mini-multileaf collimation.
Radiother Oncol 2004;71:235–41.
[13] Ma CM, Price RA, Li JS, Chen L, Wang L, Forkal E, et al.
Monitor unit calculation for Monte Carlo treatment planning.
Phys Med Biol 2004:1671–87.
[14] Mijnheer B, Battermann JJ, Wambersie A. What degree of
accuracy is required and can be achieved in photon and
neutron therapy? Radiother Oncol 1987;8:237–52.
[15] Mijnheer B, Olszewska A, Fiorino C, Hartmann G, Knöös T.A
practical guide to quality control of brachytherapy equipment.
ESTRO Booklet No. 8; 2004.
[16] NCS Report 15, Quality assurance of 3-D treatment planning
systems for external photon and electron beams: practical
guidelines for acceptance testing, commissioning and periodic
quality control of radiation therapy treatment planning
systems. Netherlands commission on radiation dosimetry;
2006.
[17] Nyholm T, Olofsson J, Ahnesjö A, Karlsson M. Photon pencil
kernel parameterisation based on beam quality index. Radiother
Oncol 2006;78:347–51.
D. Georg et al. / Radiotherapy and Oncology 85 (2007) 306–315 315
[18] Nyholm T, Olofsson J, Ahnesjö A, Karlsson M. Modelling lateral
beam quality variations in pencil beam kernel based photon
dose calculations. Phys Med Biol 2006;51:4111–8.
[19] Olofsson J, Georg D, Karisson M. A widely tested model for
head scatter influence on photon beam output. Radiother
Oncol 2003;67:225–38.
[20] Olofsson J, Nyholm T, Georg D, Ahnesjö A, Karlsson M.
Evaluation of uncertainty predictions and dose output for
model based dose calculations in photon beams from multiple
linacs. Med Phys 2006;33:2548–56.
[21] Venselaar J, Welleweerd H. Application of a test package in an
intercomparison of the photon dose calculation performance
of treatment planning systems used in a clinical setting.
Radiother Oncol 2001;60:203–13.
[22] Watanabe Y. Point dose calculations using an analytical pencil
beam kernel for IMRT plan checking. Phys Med Biol
2001:1013–38.
[23] Winkler P, Hefner A, Georg D. Dose-response characteristics of
an amorphous silicon EPID. Med Phys 2005;32:3095–105.
[24] Woo MK, Nico A. Impact of multileaf collimator leaf positioning
accuracy on intensity modulation radiation therapy quality
assurance ion chamber measurements. Med Phys
2005;32:1440–5.
[25] Yeung KT, Bortolotto K, Cosby S, Hoar M, Lederer E. Quality
assurance in radiotherapy: evaluation of errors and incidents
recorded over a 10 year period. Radiother Oncol 2005;74:283–91.