Segmentation of the myocardium from myocardial contrast echocardiography

Myocardial contrast echocardiography (MCE) is a promising new technique that allows quantification of myocardium perfusion and therefore accurate diagnosis of coronary artery disease. MCE data, however, have previously required tedious and time-consuming off-line manual image processing. This paper presents results that demonstrate success of an automatic segmentation approach utilizing active shape models. A shape model was created from a training set of eleven manually drawn contours, which was then applied to twenty-two MCE images. Standard success metrics show that error from this automatic method is comparable to error found among manually drawn contours. Additionally, a more robust calculation of the key blood flow parameters was developed which can accommodate error in the segmentation, verified by high correlation between manually and automatically derived parameters.