Automatic extraction of coronary artery tree on coronary angiograms by morphological operators

This paper presents a new method of fully automatic detection of the coronary artery tree by using several morphological filters. The contrast between coronary artery and background is enhanced by the morphological Top-Hat operation and subtraction of the result from the original image. Pixels which are obviously out of the coronary artery are removed by the half-threshold processing and the erosion operation. The labeling according to gray level difference is done on the half-threshold image so as to extract the candidates of the coronary artery. The main part of the coronary artery is extracted as the biggest area among the labeled areas. Because the tip parts of the coronary artery are narrower than the area of main coronary artery, multi-structure elements of Top-Hat operator are used repeatedly to enhance the narrow artery. The morphological watershed transform is used to detect the edge of coronary artery on the gradient image of the original image. We have tried the method on several images. Results show that the method is efficient to detect the artery tree automatically even on a less contrast image.