A Two-Stage Coronary Artery Segmentation Method Based on the Combination of 2D and 3D Convolutional Neural Networks

Cardiovascular disease is a major disease that seriously endangers public health. Coronary heart disease is the leading cause of death compared with other cardiovascular diseases. Precise coronary artery segmentation is of great significance to the treatment of coronary heart disease. Deep learning has been widely used in the field of medical imaging, but the segmentation of small object like coronary arteries is still a challenge. Aiming at accurate coronary artery segmentation this research proposes a combination of 2D and 3D convolutional neural networks. Specifically, the proposed scheme uses skeleton as a bridge to combine the two kinds of convolution networks, and expand the information receiving domain of convolution network. Compared with other deep learning based methods, the proposed method exhibits a certain improvement in sensitivity, Dice coefficient, AUC, and Hausdorf distance, and can detect the coronary arteries that cannot be identified by other competing methods, which solves the problem of vascular disconnection and blood missing to a certain extent.