Research And Implementation Of Deep Learning Algorithm For Coronary Artery Stenosis Degree Recognition

Coronary heart disease is a heart disease caused by myocardial ischemia and hypoxia,while myocardial ischemia and hypoxia are caused by coronary atherosclerosis.The report of summary of China’s cardiovascular health and disease2019 shows that the prevalence of cardiovascular diseases in China is on the rise.The number of cardiovascular patients is 330 million,of which 11 million are coronary heart disease.Interventional therapy has become an important treatment for coronary heart disease.At present,the clinical judgment of whether a patient needs interventional therapy mainly relies on coronary angiography.During the coronary angiography process,doctor manual checks the angiography images to determine the degree of coronary artery stenosis.Facing the increasing number of patients with coronary heart disease,it is undoubtedly a great pressure to rely on the doctor’s manual diagnosis,and doctors are easy to make mistakes.Therefore,based on deep learning methods,it is necessary and urgent to propose a new auxiliary method for the diagnosis of coronary artery stenosis.Due to the fast development of deep learning,scholars have proposed a large number of models for computer vision.Computer vision model can be introduced to mine medical image data to solve the problem of medical image diagnosis.Facing the diagnosis of coronary artery stenosis,a Deep Stenosis Diagnosis Model(DSDM)algorithm framework is proposed using image classification model for coronary arteries data analysis.The object detection model solves the problem of coronary artery stenosis location and the instance segmentation model solves the problem of coronary artery stenosis segmentation.With data collection,preprocessing,and annotation,a data set of the clinical diagnosis for coronary artery stenosis is constructed.Based on Retina Net model and Mask R-CNN model,an end-to-end diagnosis algorithm and a diagnosis system of coronary artery stenosis are constructed.The experimental results show that the accuracy of deep stenosis diagnosis model algorithm is 86.3% for the location of coronary artery stenosis and 86% for the segmentation of coronary artery stenosis.A coronary artery stenosis degree recognition diagnosis web service is also constructed.