Variational level set method for left ventricle segmentation

In this paper, a new variational level set method is given to segment the left ventricle walls from cine cardiac magnetic resonance images. The variational formulation consists a quasi-local binary fitting term, an area term and a penalizing term. The quasi-local binary fitting term and the area term both drive the zero active contours toward the desired object boundaries. The quasi-local binary fitting term is suitable for inhomogeneous images, and the area term can speed up the evolution of the zero level curve. The penalizing term automatically decreases the deviation of the level set function from a signed distance function, which eliminates the need of the re-initialization of the level set function.