Cardiac MR Image Segmentation with Modified Active Contour Model

This paper proposes a novel modified geometric active contour model based on level set framework and a associated scheme for segmenting the left ventricle (LV) borders from MR cardiac image sequences. By designing a new external force based on general fuzzy gradient vector flow, the algorithm improves both dynamic capture range and robustness to fuzzy object borders, and achieves successful application to the segmentation and tracking of LV borders. Experiments on clinical ECG gated 4-D cardiac MR images show that the results are very close to the manual outlines of clinical experts. LV 3D dynamic surface can be visualized, and some clinical key cardiac indices are calculated well based on the segmentation results.