ARUP; ISBN: 978-0-9562121-5-3 - CMBBE 2012 - Cardiff University

implant correctly from the finite element analysis. Despite the fact that 3D models of
any implant can be easily generated or obtained, a realistic geometry and structure for a
bone cannot be generated easily. Therefore, simplifications cannot be avoided during
the modeling of the bone. In the literature, for instance, cortical and trabecular are
mostly considered homogenous and isotropic.
Heever et al. evaluated the contact pressure of different unicompartmental implant
designs by using the finite element method [1]. In their model, the bone model was
obtained from MRI images and the material properties were assigned as cortical bone.
Baldwin et al. modeled the femur, tibia and patella by using the MRI images taken from
a cadaver and considered all the bones having only one single material properties [2].
Bougherra et al. carried out stress analysis of the bone model obtained from CT images,
which consists of two separate layers as cortical and trabecular [3]. A hybrid implant
design was developed by which the stress shielding was assumed to be reduced when
compared to traditional knee implants. Chong et al. developed a hybrid cemented tibial
implant in order to reduce the stress shielding [4]. The distal part of tibia was modeled
as isotropic while its proximal part was considered as heterogeneous during the FE
analysis.
Probabilistic approaches have been introduced recently in biomedical applications.
Nicolella et al. developed a 3D model of a cemented hip prosthesis where uncertainties
in material properties and loadings were taken into account to predict the probability of
failure [5]. Regarding probabilistic approaches used for the knee implant, Laz et al.
developed a probabilistic model of the Stanmore knee wear simulator to estimate the
variations in both the alignment of the implant and the contact pressure [6]. They
considered the loadings, component alignments as random variables affecting the knee
wear simulator. This study includes the probabilistic analysis of the simulator due to
the random variables affecting the test system. However, the analyses given in the
literature similar to this one have not considered any variability in the material
properties of the bones.
The objective of the current study was to develop a realistic patient-specific model of
knee bones from a patient’s CT images. 3D model of the knee implant as a STL file was
imported to the commercially available software package MIMICS® Medical Imaging
Software (The Materialise Group, Leuven, Belgium) in order to assembly on the bone
model. During this process, an orthopedic surgeon guided all the cuttings and assembly
procedure of the implant on the knee bones obtained from CT images. The material
properties of the bones were assigned in MIMICS based on their HU values. As
indicated in the literature, the cortical bone was assumed to be homogenous while the
material properties of the trabecular bone was assigned by grouping the elements based
on their HU values. After obtaining the implanted model of the knee bone, it was
transferred to a finite element software to carry out both stress and probabilistic analysis
to determine the performance of the knee implant under both constant and random
design variables.
3. 3D CAD MODEL OF THE KNEE IMPLANT
The CT scans of the left knee taken from a female patient at the age of 65, were
obtained from a clinical CT scanner. CT scanning parameters included: 120 kVp tube
voltage, 187.5 mAs tube current, 191 slices, 0.781 mm pixel size, 512 x 512 resolution
and 2.00 mm slice thickness. These CT images were used to obtain the 3D model of
both femur and tibia before the Total Knee Arthroplasty (TKA) surgery. Since the
patient suffers from gonarthrosis, the femur geometry as given in Fig.1 is not smooth.
2