Segmentation of MR images with B-spline snakes: A multi-resolution approach using the distance transformation for model forces

A B-spline snake algorithm is proposed for the segmentation of the knee joint cartilage from MR images. A first guess of the contour is provided interactively and is transformed into a B-spline representation. Image forces attract the B-spline curve towards the real cartilage boundaries. In addition, model forces based on a distance transformation axe introduced in order to guide the evolution of the contour in those parts of the image that show no significant features. The total energy of the B-spline curve is minimized within a multiple scale approach. The algorithm turned out to be more accurate as the manual delineation of cartilage boundaries from medical experts.