Mask R-CNN based coronary artery segmentation in coronary computed tomography angiography

Automated segmentation of the coronary artery in coronary computed tomographic angiography (CCTA) is important for clinicians in evaluating patients with coronary artery disease. Tradition visual interpretation of coronary artery stenosis exist inter-observer variability and time-consuming. The purpose of this work is to develop a deep learningbased framework for coronary artery segmentation on CCTA. We propose to use Mask R-CNN for the coronary artery segmentation. To avoid the interferences from pulmonary vessels, we propose to mask out the lung region prior to Mask R-CNN training. The network was trained using 20 patients' CCTA datasets and tested using another 5 patients' CCTA datasets. The mean of the Dice similarity coefficient (DSC) were 0.90±0.01 respectively, which demonstrated the segmentation accuracy of the proposed method.