A new approach is presented for analyzing the wall motion of cardiac left ventricle (LV) from 3D single photon emission computed tomography images. The technique applied in this paper is based on a statistical model which is one of special of physics-based deformable models. By means of content adaptive object model (CAOM) method, endocardial and epicardial surfaces of the LV can be obtained, thus a geometrical and grey-level model is built to represent wall motion by examining multi-phase cardiac images. Finally, the deformation and dynamic motion of the LV are estimated in terms of some parameters that vary during a heart cycle (about 700 ms). With the help of these parameters, physicians can characterize the motion of the LV in a clinically way and have a better treatment for possible diseases