Left ventricle segmentation from cardiac MRI combining level set methods with deep belief networks

摘要: This paper introduces a new semi-automated methodology combining level set method with top-down segmentation produced by deep belief network for the problem of left ventricle from cardiac magnetic resonance images (MRI). Our approach combines advantages that uses several priori facts about object to be segmented (e.g., smooth contour, strong edges, etc.) knowledge automatically learned manually annotated database shape and appearance segmented). The use networks is justified because its ability learn robust models few flexibility allowed us adapt it problem. We demonstrate our produces competitive results using MICCAI grand challenge on MRI images, where par best in field each one measures used (perpendicular distance, Dice metric, percentage good detections). Therefore, we conclude proposed most approaches field.