An approach for analyzing 3D cardiac motion from multi-slice computed tomographic(MSCT) angiography data is presented.First,a local vessel enhancement based on Hessian matrix and locally adaptive threshold region growing is applied to extract the 3D coronary arteries model and then 3D thinning is performed and the skeleton on different R-R intervals is obtained.The vessel branches are matched with the coherent point drift(CPD) point set registration algorithm and the correspondence matrix and transformation parameter are calculated,and thus 3D motion is estimated.The algorithm accuracy is quantitatively assessed by using simulated data with known transformation parameters,and the experiment results illustrate that the matching error is less than one voxel and the motion estimation is within 1% for a large-scale motion more than 50 mm.Real clinical full-phase data is also processed and reasonable results are achieved.