Interactive Left Ventricular Segmentation in Cardiac Images

This work presents a technique to perform segmentation of the left ventricle in images of the human myocardium. Segmentation is one of the first steps of the image analysis. Edge-based segmentation provides the detection of region boundaries, but the contours are only based on local computations, and is often sensitive to local variations in intensity, noise and physical artifacts. In interactive methods, problems like these fatigue the user and require many interventions at the contour delineation process. To reduce these problems, we propose the addition of two new features in the live-wire method: a region intensity and a proximity feature. Experimental results show that this alternative approach can achieve accurate edge localization and improved efficiency.