Coronary Heart Disease Based On Machine Learning Algorithm Risk Model And Degree Of Coronary Stenosis

Background and objective:To explore and Analysis the hazard factors of coronary heart disease(CHD)and construct CHD incident risk models based on Logistic regression,Back propagation network and XGboost from the perspective of integrated traditional Chinese and Western medicine.The aim is to select the best model to definitive diagnosis of coronary heart disease and gauge patient’s degree of coronary stenosis in a economical,efficient and uninjured way,which can actually provide scientific basis for early diagnosis,preventive intervention and definitive therapy of diseases.Methods:This study collected patients who have already perfected coronary angiography(CA)in tcm-Integrated Hospital of Southern Medical University from October 2019 to September 2021.According to the questionnaire survey and medical record file,collect the basic information,physical examination,comorbidities,Chinese medicine syndrome and inspection results.The content studied consists of three parts:(1)Using statistical methods to explore and select the specific and differential risk factors of CHD.(2)Constracting Logistic regression,Back propagation network and XGboost model associated with CHD and evaluating them by using quantitative indicators such as accuracy,specificity,recall and ROC curve to select the most appropriate model.(3)Building precise model with machine learning to identify the degree of coronary stenosis,including LAD,LCX and RCA.Results:1.A total of 256 patients were included in the the study,and the morbidity rate of coronary heart disease in the research data was 62.11%,63.35%for male and 43.16%for female.2.The model based on Xgboost algorithm has the best performance of distinguishing the CHD,with high accuracy(0.969),specificity(0.968)and ROC curve(0.969).As for the structure of predictive model,the most important and the most relevant risk factors of CHD are age,sex,LDL,HbA1c,ATPP,FIB,WBC,MYO and cTnT.AND the predictive models used to identify the degree of LAD,LCX,RCA also have good prediction ability,all of them have good accuracy and specificity,and the MEA of coronary artery(LAD,LCX,RCA)are 11.26%,11.32%and 9.33%,the R2 are 0.79,0.71 and 0.79.The Common predictors of such three models are age,sex,LDL,ATPP,FIB,WBC,MYO and cTnT.Conclusions:Among patients in the research data,the prevalence of CHD is extremely high,and the age of onset assumes the youth oriented tendency.As for the predictors and risk factors,most of the physical,chemical indicators and Chinese medicine syndrome play important roles in the predictive model,and above of them have significant statistical difference between CHD and non-CHD patients.In the study,both of the classification model and classification model demonstrate extremely high performance,their AUC values and accuracy values reflect the high performance of the model.The models are proven as Effective methods in clinical practice of preventing coronary heart disease.