Segmentation of the Left Ventricle from Cardiac MR Images Based on Degenerated Minimal Surface Diffusion and Shape Priors

Abstract

Segmentation of cardiac MR images is a hot topic in the community of medical image analysis and remains one of the open problems. In this study, we address the segmentation of 2D cardiac MR images based on gradient vector flow snake and make three contributions: firstly, the degenerated minimal surface GVF is proposed, this new flow outperforms the original one referring to boundary preserving; secondly, the shape of the left ventricle is taken into account and a shape based energy for the snake model is adopted, with this energy, the snake contour can conquer the unexpected local minimum stemming from image inhomogeneity and the final results could depend much less on the initial contour. In order to extract the epicardium, a novel approach is developed to generate the desired external force. This new external forces can overcome the demerits of the GVF force directly derived from the usual edge map and maintains the epicardium boundaries even if the contrast between the myocardium and neighbor organs is very low, taking the endocardium contour as initialization, the snake can capture the epicardium fast and accurately. The proposed strategy is also validated experimentally.