Direct measurement of cartilage contact is not feasible, necessitating the use of musculoskeletal models to estimate the internal soft tissue loading associated with movement2. However, current computational models used to simulate movement often utilize simplified joint kinematic constraints which can only provide an estimate of the net joint reaction force1. Detailed finite element models of the knee have been created14 which can provide estimates of cartilage tissue stress, but are too computationally expensive to solve within the context of a whole body simulation of movement. Discrete element analysis (DEA) provides a viable alternative for rapidly computing contact stress patterns in movement15, However, even DEA models can be computationally expensive as high resolution polygonal meshes are needed to accurately represent complex cartilage geometries such as that seen on the femoral condyles and tibia plateau. In a DEA model, much of the computation time is spent querying for contact between triangles of two polygonal surfaces. The objective of this study was to investigate the potential for using graphic processor computation (GPU) to expedite contact detection5. To do this, we use a contact detection algorithm that pre-constructs hierarchical bounding-volumes (BVH) of the target body to increase the efficiency of contact detection. We then show that implementing a parallel computational version of this algorithm on the GPU greatly speed up performance and thus make it more viable to simulate cartilage contact within the context of human movement.

This content is only available via PDF.
You do not currently have access to this content.