One-click coronary tree segmentation in CT angiographic images

In this paper we present a method to automatically detect the coronary arteries in CT angiographic images after a single user click into the aorta. The algorithm consists of three steps: the masking of the aorta, the analysis of connected regions outside the aorta, and the vessel segmentation starting from the detected seed clusters. The aorta segmentation and the detection of voxel clusters at the coronary artery origins is based on a statistical analysis of regions segmented by a 2D region growing algorithm. The method has been tested with 61 data sets from different sites and different CT scanners including cases with abnormal anatomy or bad image quality Reasonable results were produced in all cases. In 57% of all cases, both coronary arteries were detected, while in 85% at least one coronary artery could be found. As the results can be easily extended manually by mouse clicks into arteries that were not detected automatically the presented method is appropriate as a first step in an advanced analysis of the coronary tree.