Currently, we solve the boundary finding problem twice for each 2D image slice of a 3D image frame, once for the endocardial border and another for the epicardial one, using the parametric deformable contour model approach developed in our laboratory[38]. The boundaries found in one time frame are used as a bias and initial estimate for locating the borders in the next time frame. This process repeats until all of the contours that make up the LV surfaces in each frame are detected.