Fast and automatic coronary artery segmentation using nnU-Net for non-contrast enhanced magnetic resonance coronary angiography

Motivation: Non-contrast MRCA is promising for coronary heart disease screening, but its low spatial resolution and poor contrast between coronary arteries and surrounding tissues make automatic segmentation challenging. Goal(s): we introduced a deep learning model, nnU-Net, for MRCA image segmentation Approach: The nnU-Net was trained on 134 datasets and evaluated on 114 others. Two radiologists qualitatively assessed the segmented arteries as good to excellent.and compared with CCTA data, showing comparable results in terms of branch numbers, total branch length, and distance from the base to the main coronary artery origin. Results: experiment results demontated that nnU-Net can accurately extract the coronary arteries from MRCA. Impact: The coronary ateries can be correctly segmented from MRCA by nnU-Net. MRCA combined with nnU-Net can be an screening tool for coronary heart disease in clinical practice.