An automated method for simultaneous detection of left and right coronary borders

Summary form only given. An automated method for simultaneous detection of both coronary borders from angiograms is developed that produces detected borders that are optimal when taken as a pair. In this method, a two-dimensional edge operator was applied to cine frames digitized at 3* into a 512*512*8 matrix. The resulting edge images were resampled perpendicular to the vessel center-line and the resampled profiles were used to construct a three-dimensional directed graph. An optimal path through the graph, representing a pair of coronary borders, was found using a minimum cost search algorithm. Coronary borders detected by the method were visually compared by an observer who was unaware of the method of analysis. In the set of 19 'difficult' images, the method yielded superior results in 8 of 19 images, equivalent results in 10 images, and a poorer result in 1 image.<
>