Fast detection & modelling of the real osteoarthritic holes in the human knee with contour interpolated radial basis functions

In this article, we propose a novel method for the fast 3D reconstruction of real osteoarthritic (OA) holes in a human femoral cartilage. Initially, semi-automated Region-Based Segmentation (region-growing) and Bounding Box techniques are used to extract femoral cartilage slices from MRI scans of the knee. OA holes were detected and filled automatically by our contour interpolation/RBF (CI-RBF) method and 3D models of both the femoral cartilage and OA holes were reconstructed separately. The method was then applied to a single human knee and results proved it fast, reliable and accurate for reconstructing a 3D model of the femoral cartilage from MRI images with an extremely low root mean square error of 1.67% in the estimated volume of the automatically filled to the manually filled femoral cartilage slices. As per authors´ knowledge this is the first time real OA hole has automatically been identified and filled.