A new algorithm to detect irregular coronary boundaries: the gradient field transform

A new algorithm is described that was specially designed for the detection of sharply and abruptly changing vessel dimensions as in complex lesions. The algorithm, the gradient field transform, uses the directional gray value information available in digital arteriograms. The search strategy, which is based on Dijkstra's algorithm, finds a minimal cost edge along a coronary artery, even those with very irregular boundaries. Measurements on copper obstructions with very short lengths demonstrated an overestimation of less than 0.2 mm, while the more conventional minimum cost algorithm overestimated the obstruction diameters by 4 mm.<
>