Radiomics Model For Evaluating Myocardial Ischemia In Coronary CT Angiography

ObjectiveTo investigate the feasibility of radiomics model for prediction of myocardium ischemia in CCTA images.MethodsSuspected CAD patients from March 2011 to December 2015 who underwent CCTA and SPECT within 3 months were retrospectively enrolled.Ischemic and nonischemic segments were confirmed by SPECT as reference standard and were randomly assigned as primary and validation set at the ratio of 3:1.Myocardium segments in CCTA images were divided based on AHA 17-segment model.Segmentation and Feature extraction were implemented in Cardiac Function Analysis and Radiomics respectively.Least Absolute Shrinkage and Selection Operator(LASSO)regression was used for data dimension reduction and feature selection in the primary set.Multivariable logistic regression analysis was used to develop the predicting model.ROC curve analysis was performed to evaluate the performance of prediction model containing sensitivity,specificity and accuracy.Results89 patients containing 374 non-ischemic and 151 ischemic segments were enrolled,and 17 ischemia status-related with nonzero coefficients features were selected for further analysis.The prediction model showed good discrimination in both primary dataset(AUC=0.81;sensitivity=73.9%,specificity=73.0%,accuracy=73.4%)and validation dataset(AUC=0.75,sensitivity=67.7%,specificity=78.3%,accuracy=71.9%).ConclusionIschemic and non-ischemic myocardium differ in radiomics features.It’s feasible to establish a radiomics model from CCTA images to discriminate the myocardial ischemia,but its accuracy needs to be improved.