In this paper, we describe a method for automatically building a statistical shape model by applying a morphing method and a principal component analysis (PCA) to a large database of femurs. One of the major challenges in building a shape model from a training data set of 3D objects is the determination of the correspondence between different shapes. In our work, we solve this problem by using a morphing method. The morphing method consists of deforming the same template mesh over a large database of femur geometries, which results in isotopological meshes and one to one correspondences; i.e., the resulting meshes have the same number of nodes, the same number of elements, and the same connectivity in all morphed meshes. By applying the morphing-based registration followed by PCA to a large database of femurs, we demonstrate that the method can be used to derive a low dimensional representation of the main variabilities of the femur geometry.
Combining statistical shape model and principal component analysis to estimate left ventricular volume and ejection fraction
