QUANTITATIVE CORONARY ANGIOGRAPHY WITH DEFORMABLE SPLINE MODELS

Although current edge-following schemes can be very efficient in deter mining coronary boundaries, they may fail when the feature to be follo wed is disconnected (and the scheme is unable to bridge the discontinu ity) or branch points exist where the best path to follow is indetermi nate. In this paper, we present new deformable spline algorithms for d etermining vessel boundaries, and enhancing their centerline features, A bank of even and odd S-Gabor filter pairs of different orientations are convolved with vascular images in order to create an external sna ke energy field, Each filter pair will give maximum response to the se gment of vessel having the same orientation as the filters. The result ing responses across filters of different orientations are combined to create an external energy field for snake optimization, Vessels are r epresented by B-Spline snakes, and are optimized on filter outputs wit h dynamic programming. The points of minima! constriction and the perc ent-diameter stenosis are determined from a computed vessel centerline . The system has been statistically validated using fixed stenosis and flexible-tube phantoms. It has also been validated on 20 coronary les ions with two independent operators, and has been tested for interoper ator and intraoperator variability and reproducibility. The system has been found to be specially robust in complex images involving vessel branchings and incomplete contrast filling.