Spatio-temporal grouping for the formation of vascular segments in coronarography image sequences

This work is aimed at the extraction of vascular structures using perceptual grouping methods. The approach is based on structural feature matching, between images in the time sequence. These features correspond to junctions (or bifurcation points) and are paired by means of a local criteria. A global refinement procedure is then applied by means of a relaxation scheme. The feature trajectories over the entire image sequence are examined to further build sets of vascular segments, which better describe vascular branches. The resulting performances are exemplified on a standard biplane coronarographic examination.