Experimental Study On Prediction Of Major Adverse Cardiovascular Events After Acute Myocardial Infarction Based On Machine Learning

Acute myocardial infarction(AMI)is a kind of disease caused by acute occlusion of coronary artery caused by various reasons,which leads to continuous ischemia and hypoxia of heart and myocardial necrosis.The disease has the characteristics of acute onset,rapid development,high incidence and mortality.At present,the most critical treatment is to open blood vessels as soon as possible and restore blood and oxygen supply to the heart.However,a large number of patients will still have major adverse cardiovascular events(MACEs)after active revascularization,including myocardial infarction,heart failure,renal failure,coronary events,cerebrovascular events and death.Therefore,the prediction and evaluation of MACEs after AMI and individualized treatment according to its incidence are of great significance to improve the prognosis of patients with AMI.The traditional risk assessment and prediction models have small amount of data,simple structure and low accuracy,which is difficult to effectively predict MACEs after AMI.Therefore,this thesis carried out an experimental study on the prediction of MACEs after AMI based on machine learning,and the main research work is as follows:(1)Based on the structured electronic medical record data of AMI patients,this topic first used the binary logistic regression method to analyze the independent risk factors of MACEs after AMI;then seven machine learning models(including logistic regression,decision tree,naive bayes,support vector machine,random decision forest,gradient boosting and multilayer perceptron)were trained and tested,and the prediction performance of each model on MACEs after AMI was evaluated.This study found that cardiac function grade,drug compliance,age,creatinine and total cholesterol levels were independent predictors of MACEs after AMI;it was verified that the random forest model was the best model for risk stratification of AMI patients.(2)Based on the unstructured coronary angiography data of AMI patients,this paper first optimized the input size and spatial pyramid pooling module of YOLO v4,and used it to detect meaningful stenosis,diffuse stenosis and chronic total occlusion.Secondly,the value of three lesions in the evaluation of MACEs after AMI was studied.The results showed that the mean average precision of the improved YOLO v4 model was 55.01%,and chronic total occlusion is of great significance in the evaluation of MACEs after AMI.In conclusion,the random forest algorithm has high accuracy in predicting MACEs after AMI;YOLO v4 algorithm can effectively identify coronary artery lesions.Combining the structured and unstructured data of AMI patients to build a machine learning model can predict and evaluate MACEs after AMI more intelligently,efficiently and accurately.