Distribution Matching and Active Contour Model based Cardiac MRI Segmentation

摘要

In this paper an approach to extract the Left Ventricle (LV) endocardium contour by proposing an improved Distribution Matching (DM) algorithm is presented, the main idea of which is to match the distribution of grayscale and distance constrained by gradient vector flow (GVF) force field between the sample and the images input. The endocardium contour is then regarded as the initial iterative curve of Active Contour Model (ACM) to obtain the epicardium contour with a method of local circular constraints and adaptive changing parameter of external force field. The approach has been applied on data of 20 subjects, and results have demonstrated simpler user interaction, more robustness and higher segmentation accuracy compared to LV segmentation methods published previously.