A novel method for segmentation of the cardiac MR images using generalized DDGVF snake models with shape priors

In this study, a novel method is presented for segmentation of the endocardium and epicardium of the left ventricle in cardiac magnetic resonance images using snake models. We first generalize the DDGVF snake model by introducing two spatially varying weighting functions which characterize the boundary information: this generalized DDGVF snake can conquer the spurious edges raised by artifacts while maintaining the desirable properties of DDGVF of distinguishing the positive and negative boundaries. This is especially helpful for the tasks on hand because the endocardium and epicardium of the LV in MR images can be characterized as positive and negative boundaries. Observed that the left ventricle is roughly a circle, a shape constraint based on circle is introduced into the snake model. This new constraint can prevent the snake contour from being trapped and leaking out so as to maintain the global shape of the snake contour during evolution. In addition, fourth-order PDEs are employed for noise removal. We demonstrate the proposed approach on an in vivo datasct and compare the segmented contours with manual collections; the results show its effectiveness.