Segmentation and Tracking of Coronary Artery Using Graph-Cut in CT Angiographic

This paper presents a semiautomatic method to detect the coronary arteries in X-ray CT angiography (CTA) images with simple user interaction. The algorithm is started with performing modified Otsu threshold method and clustering of superpixel to reconstruct the image. The interactive graph cut segmentation and an improved top-hat filter is used to extract the coronary artery in series of two dimension (2D) slice images. Finally, coronary artery tree is tracked with a method based on region average shift. The accuracy of coronary artery extraction has been validated in 16 CTA datasets. The results demonstrate this algorithm has satisfactory speed and efficiency. With more user interaction, the method can be easily extended to the arteries that are not detected automatically.