Model-based segmentation of the left main coronary bifurcation from 2D angiograms

This paper deals with the identification of the left main coronary bifurcation (LMB) in 2D angiograms. We resort to a generic 3D model of the LMB that is projected in the proper angulation to retrieve the expected 2D skeleton of the LMB in the considered image. We derive from it an efficient discriminating filter that allows detecting the LMB, while being robust to anatomical variations. We demonstrate the potential of the method over a large database of 120 angiographic sequences from 15 patients. We also propose a method extension that exploits not only the present angiography, but also the previous ones, as acquired in the course of the exam.