Detecting Abnormality on Coronary Artery Image by Extracted Edges to Deep Learning

Abnormalities in the coronary arteries are one of the causes of Coronary Artery Disease (CAD), which is the most harmful fatal disease in the world. To diagnose CAD, it requires a good doctor to have a lot of experience as well as a lot of time to consider. We found that after coronary segmentation, the coronary edges would not be good to identify abnormalities. We propose to improve by extracting coronary artery edges and then using the vessel wall browsing algorithm to locate abnormalities on the blood vessels [1], this improved algorithm reached 74.9% (it is better than the original method is 71.4%). The second method, the deep learning method, we apply a convolutional neural network (CNN) model to classify a coronary image is normal or abnormal. Experiment results from our private dataset show that our methods have an accuracy of 75.4%.