Automatic cardiac contour propagation in short axis cardiac MR images

Active contours are a popular method for automatic extraction of object boundaries based on image features in medical images. However, in short axis cardiac MR images, they fail to give correct results due to the presence of papillary muscles along the left ventricular endocardium boundary. We propose a new automatic cardiac contour propagation method based on active contours. The method can be used to propagate cardiac contours that conform an initial manual segmentation, by exploiting information in adjacent images. In this paper, we present the results of the optimization and extensive validation of our method, which was found to be robust and accurate for the delineation of the LV endocardium, LV epicardium and RV endocardium contours. Additionally, the processing time is reduced significantly compared with manual contour delineation.