Improving Segmentation of the Left Ventricle Using a Two-Component Statistical Model