Cardiac MR image segmentation: quality assessment of STACS

We present an energy based automatic image segmentation algorithm that uses a novel active contour scheme, called the stochastic active contour scheme (STACS). The algorithm overcomes some unique challenges arising in cardiac magnetic resonance (MR) images by minimizing an energy functional with four terms, each representing the region and edge based information of the image and the global and local properties of the contour. We use annealing schedules to control the relative strength of each of the terms during the minimization process. The segmentation results when applying STACS to a set of real cardiac MR sequences of a rat are presented and quantitatively assessed by comparing them to the manually-traced contours using two similarity measures, the area and shape similarity measures. This assessment validates STACS's results, demonstrating its very good and consistent segmentation performance.