Fast Subject Specific Finite Element Mesh Generation of Knee Joint from Biplanar X-ray Images

Numerous finite element (FE) models of the knee joint have been developed to investigate knee pathology, post-surgery assessment and natural knee biomechanics. However, because of the extensive computational effort required for preparing subject specific model from CT-scan or MRI data, most of the models in literature are done only for one subject resulting in poor validation of the model and limits the predictive power of the conclusions. Biplanar X-ray is a promising alternative to perform 3D reconstruction of bony structures because of low radiation dose and very less reconstruction time [1]. Moreover, an accurate and fast computational mesh is a prerequisite for generating subject specific mesh in order to perform personalized FE analysis. Traditionally, both triangular/tetrahedral and quadrilateral/hexahedral FE elements are used for 3D mesh generation. But because of distinct numerical advantages quadrilateral/hexahedral elements are preferred to avoid numerical instability, specifically for problems involving high strains at soft tissues [2]. The aim of the current study is to develop fast and automatic subject specific mesh for knee joint from biplanar X-ray images. This approach was successfully tested for 6 cadaveric specimen, where from the biplanar radiographic images of each, 3D reconstruction models were built with a mean time of about 10 min for each specimen by adapting the strategy of [1]. From the reconstruction models, subject specific mesh (4 noded shell) for bony and cartilage structures were generated based on the mapping from the generic model to subject specific model with about 10 sec of time for each specimen (Fig. 1). Both the meniscus were meshed with 8 noded hex elements using the nodes of femoral and tibial cartilage in a dedicated Matlab code with numerical cost of almost 1 min. So, a total of about 12 min computational time was required to build each subject specific knee from 3D reconstruction to mesh generation which is promising for clinical applications. Quality of mesh for individual specimen was also checked using mesh quality indicators (Jacobian ratio, aspect ratio etc.) and surface representation accuracy, which showed less than 1% (warning only) and 0.8 mm (at soft tissue regions) respectively for individual specimen