Objective To explore the application value of machine learning models in predicting acute coronary syndrome based on clinical indicators and pericoronary adipose tissue(PCAT)radiomics features in coronary computed tomography angiography(CCTA)of patients with coronary heart disease(CHD).Methods The clinical data and CCTA of 164 patients with CHD were analyzed retrospectively.Based on medical record follow-up,the patients were divided into two group:stable coronary artery disease(SCAD)group and acute coronary syndrome(ACS)group.The patients were randomly divided into a training group and a validation group in the ratio of 7∶3.The region of interest for PCAT was delineated on CCTA images.Radiomics features were extracted and selected by least absolute shrinkage and selection operator to build the radiomics model.By integrating the clinical data and radiomics features,the predictive models were developed using five classifiers,including decision tree,extreme gradient boosting,support vector machine,random forest and logistic regression(LR).The predictive performance of the models was assessed by using the receiver operating characteristic curve,with differences compared by using DeLong test.Calibration curves were used to evaluate model accuracy,and the clinical benefits were assessed by decision curve analysis.Results Of 164 patients with CHD,there were 107 SCAD and 57 ACS.According to the univariate and multivariate logistic regression analysis,the history of diabetes mellitus and CT-derived fractional flow reserve were the independent risk factors of ACS(P＜0.05).Among all models,the clinical-radiomics model based on LR showed the highest predictive performance with area under curves of 0.951(95％CI:0.912,0.990)in the training set and 0.845(95％CI:0.713,0.977)in the validation set,significantly outperforming other classifiers(P＜0.05).Calibration curves revealed good model accuracy,and decision curves demonstrated high clinical benefit for the combined clinical-radiomics model.Conclusion The machine learning model based on CCTA pericardial adipose tissue radiomics and clinical features can effectively predict the risk of ACS in CHD.
维普
万方数据