Markov Shape Models: Object Boundary Identification in Serial Magnetic Resonance Images
Underhill H, Kerwin W
University of Washington
A method to perform object boundary identification in sequential images is presented. Building on the structure of the Active Shape Model, the proposed technique utilizes object descriptors from the previous image to refine the search in the current image. The underlying idea is similar to Markov random processes and is therefore referred to as a �Markov Shape Model.� The method is successfully applied to outer-wall boundary detection of the common carotid artery and its highly variable bifurcation in a sequence of magnetic resonance images (MRIs). The technique may be an effective tool for automated object boundary detection in MRI.