Segmenting a beating heart using polysegment and Spatial GPCA

Given a volume of cardiac MR images, we consider the problem of segmenting the heart based on intensity and dynamics. We first segment the heart and the chest from the background using an algebraic technique for intensity based segmentation called polysegment. As the heart and the chest exhibit different dynamics, we model the image temporal evolution as the output of two different linear dynamical systems. Under this model, the trajectories of the heart and chest intensities lie in different subspaces. We thus propose a method called Spatial GPCA for clustering data points lying in multiple subspaces, while maintaining the spatial coherence of the data points. We compare the segmentation results of Spatial GPCA to those of K-subspaces, another popular subspace clustering algorithm