Development of subject-specific and statistical shape models of the knee using an efficient segmentation and mesh-morphing approach
Introduction
Subject-specific finite element (FE) models incorporating anatomical articular cartilage surfaces and soft tissue geometric representations can provide insight into knee mechanics for healthy normal and pathologic conditions [1], [2], [3]. Accurate prediction of knee joint mechanics in FE models depends on multiple factors including appropriate representations of the geometry, properties of anatomic structures and the application of boundary conditions (i.e. kinematics and muscle forces) [4]. Segmentation of computed tomography (CT) and magnetic resonance (MR) scan data has become the accepted standard for subject-specific model development [4], [5], [6]. However, the processing is typically manual and time-consuming; extracting the articular surfaces of a knee joint was reported as requiring approximately 2 days of work [6]. In addition, model development time is increased with the meshing of segmented surfaces into 3D hexahedral solids, which are recommended for accurate FE representations for bone strain [7] and contact [8]. The hexahedral meshing process is also manual, requiring user interaction and advanced knowledge of mesh generation techniques [9].
Recently, statistical shape models have been demonstrated for bony geometries [10], [11], [12], [13] with the potential to efficiently generate a patient-specific model from anatomical measurements for use in computer-assisted surgery or to create a simulated population of subjects for assessment of implant design. Statistical shape modeling involves a principal component analysis (PCA) performed on a training set of extracted subject geometries to determine the modes of spatial variation [10], [11], [12], [13]. In addition to manually segmenting the geometries to form the training set, a traditional challenge in statistical shape modeling is that registration, a correspondence of the landmark locations (mesh nodes) for each instance in the training set, requires implementation of custom algorithms and can be computationally expensive [10], [11], [12], [13].
To address these issues of model development efficiency, recent studies have investigated various aspects of automating the model development process. Automated threshold-based algorithms have been employed to extract bones from CT scans [14], [15], however, these techniques have rarely been applied to soft tissue structures (e.g. cartilage, ligaments, muscles) from MR scans [16]. Another common approach is to use traditional segmentation of CT images to generate a ‘target’ surface and to automate the mapping of a template mesh to fit the subject-specific segmented surfaces [17], [18], [19], [20], [21], [22]. Additionally, deformable image registration techniques have been applied to noninvasively measure strain in knee ligaments [23].
The current study enhanced model development efficiency by proposing an integrated segmentation and hex meshing approach that is applicable to structures in both CT and MR scans, has accuracy similar to manual segmentation techniques, and can facilitate statistical shape modeling. As each of the subject models will have the same underlying mesh, the control points can be used directly in the formulation of a statistical shape model. Accordingly, the objectives of the current study were: (1) to develop an efficient, integrated mesh-morphing-based segmentation approach to create hexahedral meshes of subject-specific geometries from scan data and to apply the approach to natural femoral, tibial, and patellar cartilage (and the patella) from MR images, (2) to compare geometries and predicted contact results from a quasi-static FE analyses between meshed surfaces using the semi-automated approach and traditional segmentation, and (3) to demonstrate the direct application of the subject-specific models to create a statistical shape model of the knee characterizing the modes of variation using PCA.
Section snippets
Integrated segmentation and mesh-morphing platform
The integrated platform utilized a custom graphical user interface (GUI) and required a template mesh for each structure. The template mesh, developed using the traditional approach of manually segmenting (ScanIP, Simpleware, Exeter, UK) and meshing the structures of interest for a single subject, was used for morphing the geometries of subsequent subjects. Mesh morphing of the template was conducted using built-in Hypermesh (Altair, Inc., Troy, MI) features that maintained internal element
Results
Hexahedral meshes of femoral, tibial, and patellar cartilage were successfully generated using the integrated segmentation, mesh-morphing approach on MR data from 10 subjects. Accuracy of the extracted geometry was evaluated by comparing geometric differences between the semi-automated and traditional manual approaches for three subjects; average RMS differences across all articular geometries were 0.54 ± 0.32 (S.D.) mm and are reported in Table 2 for each structure and subject. Creating a
Discussion
Clinically relevant issues involving the natural knee, such as tibiofemoral osteoarthritis and patellofemoral pain syndrome, continue to warrant biomechanical evaluations to understand their etiology and potential intervention strategies [4]. The FE method has proven useful in understanding natural knee kinematics, contact mechanics, and internal stresses and strains [1], [3], [27], but the extensive time required to generate specimen-specific hexahedral meshed cartilage structures, required
Conflict of interest
There are no conflicts of interest.
Acknowledgements
The authors would like to thank Dr. Lorin Maletsky and Chadd Clary at the University of Kansas for providing image data. The authors acknowledge the assistance of Cameron Lemmon and Rakesh Ramachandran with segmentation. This research was supported in part by DePuy, a Johnson & Johnson Company.
References (34)
- et al.
A three-dimensional finite element analysis of the combined behavior of ligaments and menisci in the healthy human knee joint
Journal of Biomechanics
(2006) - et al.
The effect of graft stiffness on knee joint biomechanics after ACL reconstruction—a 3D computational simulation
Clinical Biomechanics
(2003) Inter-rater and intra-rater reliability of subchondral bone and cartilage thickness measurement from MRI
Magnetic Resonance Imaging
(2003)- et al.
Knee cartilage topography, thickness, and contact areas from MRI: in-vitro calibration and in-vivo measurements
Osteoarthritis and Cartilage
(1999) - et al.
Comparative study on different methods of automatic mesh generation of human femurs
Medical Engineering & Physics
(1998) - et al.
Statistical deformable bone models for robust 3D surface extrapolation from sparse data
Medical Image Analysis
(2007) - et al.
Instantiation and registration of statistical shape models of the femur and pelvis using 3D ultrasound imaging
Medical Image Analysis
(2008) - et al.
Subject-specific finite element models of long bones: an in vitro evaluation of the overall accuracy
Journal of Biomechanics
(2006) - et al.
Border-tracing algorithm implementation for the femoral geometry reconstruction
Computer Methods and Programs in Biomedicine
(2001) - et al.
The mesh-matching algorithm: an automatic 3D mesh generator for finite element structures
Journal of Biomechanics
(2000)
Automated three-dimensional finite element modelling of bone: a new method
Journal of Biomedical Engineering
The use of sparse CT datasets for auto-generating accurate FE models of the femur and pelvis
Journal of Biomechanics
Mesh-morphing algorithms for specimen-specific finite element modeling
Journal of Biomechanics
Knee joint biomechanics in open-kinetic-chain flexion exercises
Clinical Biomechanics
Patellar contact forces with and without patellar resurfacing in total knee arthroplasty
Journal of Arthroplasty
Knee joint mechanics under quadriceps—hamstrings muscle forces are influenced by tibial restraint
Clinical Biomechanics
Automatic generation of accurate subject-specific bone finite element models to be used in clinical studies
Journal of Biomechanics
Cited by (98)
Automated 2D and 3D finite element overclosure adjustment and mesh morphing using generalized regression neural networks
2024, Medical Engineering and PhysicsStatistical modeling: Assessing the anatomic variability of knee joint space width
2023, Journal of BiomechanicsTransformer-based multilevel region and edge aggregation network for magnetic resonance image segmentation
2023, Computers in Biology and MedicinePersonalised statistical modelling of soft tissue structures in the ankle
2022, Computer Methods and Programs in BiomedicineReporting checklist for verification and validation of finite element analysis in orthopedic and trauma biomechanics
2021, Medical Engineering and Physics