Multiframe estimation of contour evolution in medical images

A framework for temporal estimation of contour deformation in image sequences has been developed. The proposed solution for point-wise tracking finds, for a point, one optimum trajectory across multiple frames, in contrast with frame-to-frame tracking (2-frame). The multiframe approach aims (i) to find the complete point trajectory as the global optimal track over the space with multiple contours, and not as optimal sub-tracks concatenated over this set of contours; and (ii) take into account information from an extended time interval. The multiframe solution is framed as a shortest-path problem and uses the Dijkstra´s algorithm. The method applies the dynamic programming technique to find the global trajectory. The algorithm requires no special initialization or markers. It is a non-invasive method and image modality independent. The usefulness of the multiframe tracking is illustrated using synthetic images. The method demonstrated to be consistent with methods published in the literature.