Efficient and accurate femur reconstruction using model-based segmentation and superquadric shapes

In this paper, we present a new method for the segmentation of human bones in MRI (Magnet Resonance Imaging) scan images. To illustrate the approach, we focus on the modeling of a real femur. Region growing and threshold methods are used for pre-processing. For some areas of bones, for example, the femur ball, we use a VRML (Virtual Reality Modeling Language)-based femur model as a reference in order to make the segmentation method more robust in its response to noise and the influence of other tissues with the same grey level as bones. In parallel, we propose an accurate and efficient geometric fitting and reconstruction model using a composite superquadric (SQ) model. The accuracy of the algorithm is demonstrated through the reconstruction of a femur dataset and further functional parameters in comparison with manual extraction and the VRML-based approach.