Left ventricle tracking in isotopic ventriculography using statistical deformable models

The left ventricle tracking in planar scintigraphic images is an important step to study the coronary state. However, it is not as simple as we thought, because of the bad quality of these images. The study of the related work showed that classic methods based on low level image characteristics failed to localize the ventricle. In this paper, we suggest using a method based on a priori knowledge. Assuming that the human heart can appear in different images as one of a variety of possible shapes, we can incorporate the shape and the texture to delineate the boundaries of the left ventricle.