Segmentation of the Full Myocardium in Echocardiography using Constrained Level-Sets

In echocardiography, left ventricle detection is a common practice in order to retrieve indexes of myocardial health. Although a great attention has been given to the segmentation of the endocardium, very limited literature addresses the detection of both endo- and epicardial contours. Hereto, in this study we propose an original levelset technique specifically designed for the detection of the whole myocardium in short-axis ultrasound scans. A localized version of standard region-based methods is adopted, and allows to deal with low-contrast boundaries. Moreover, shape prior information is efficiently embedded in the evolution equation, thereby preventing the detection of undesired structures, like papillary muscles. An evaluation is presented on a set of 40 frames from 5 different patients. The proposed method is shown to be a reliable tool for an accurate segmentation of the full myocardium on ultrasound scans.