Motion estimation for sequential medical images using a deformable model

The gradient method, which is a conventional method of motion analysis, has the problem that artifacts appear near contours or edges with a large brightness gradient. Another method, which applies a deformable model to the deforming object, has the opposite problem, namely, that the accuracy of estimation is degraded in areas with small brightness gradients. In order to handle such problems, this paper proposes a new deformable model in which the motion of each part of a pliantly deforming object is estimated. In this method, a term based on local optimization is added to the interframe energy following the conventional definition, so that the shape and the motion minimizing the energy are simultaneously determined. More accurate estimation of the motion is realized owing to the multiplicative effect of the term based on the brightness gradient composing the interframe external energy, and the term based on local optimization. The proposed method is applied to time-series chest MR images and its effectiveness is verified. © 2005 Wiley Periodicals, Inc. Syst Comp Jpn, 36(4): 27–36, 2005; Published online in Wiley InterScience (). DOI 10.1002sscj.20146