An automatic and efficient coronary arteries extraction method in CT angiographies. (July 2017)

• Record Type: Journal Article
• Authors: Li, Zhixun; Zhang, Yingtao; Gong, Huiling; Liu, Guangzhong; Li, Weimin; Tang, Xianglong
• Is Part Of: Biomedical signal processing and control. Volume 36(2017)
• Page Start: 221
• Page End: 233
• Publication Date: 2017-07
• Subjects: Coronary CT angiography -- Coronary artery detection -- Heart region isolation -- Vessel identification -- Centerline extraction -- Vessel connectedness -- Principal curve; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2017.04.002
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A fully automatic vascular centerline extraction method is proposed in CT angiography. We does a DBSCAN-based method to isolate the heart region. Anisotropic distribution is employed to discriminate the real vessel and false response. We identify the shape of blood vessels via connectedness-related clustering. The method can obtain good performance for the vascular centerline extraction. Abstract: Coronary artery disease (CAD) is one of the most leading cause of death in the world recently. Although conventional coronary angiography (CCA) is still as a gold standard for diagnosis of CAD, coronary CT angiography (CTA) is widely used due to its non-invasion. Meanwhile, centerline extraction of the coronary arteries offers some essential information for the radiologists. However, their extractions are extremely difficult due to the CT imaging nature, such as strong motion artifacts, poor contrast injection timing and low contrast. In addition, as the data increases rapidly, manual extraction becomes impractical, a fully automatic extraction method is becoming more necessary. In this paper, we propose an automatic and efficient method to extract the vascular centerlines in CTA. Firstly, the heart region is isolated through chest wall and spine removal based on CT thresholding and DBSCAN clustering. Secondly, real coronary arteries are distinguished from artifacts via an identification function based on their different anisotropic distributions of Frangi's vesselness. And good heart isolation and artifact removal can avoid the detection of the aorta and ostium always used in other works. Thirdly, we study a directional connectedness-related clustering method to cluster every vascular segments for forming a reasonable vascular tree, as well as our method has ability at collateral vessel's elimination in refinement step according to actual demand. Fourthly, in order to handle some break-offs and holes in post-enhanced image, centerlines are extracted by employing a robust method based on principle curves. Finally, the performance of our method is evaluated on RCAA datasets in CTA. Our method performs well followed by a comparison of the-state-of-the-art methods.