| ObjectiveTo investigate the value of establishing a normogram for predicting coronary artery disease(CAD)with coronary computed tomographic angiography(CCTA)-based radiomic signature of pericoronary adipose tissue(PCAT)combined with clinical indicators.Materials and MethodsA total of 157 patients who underwent CCTA for suspected CAD and invasive coronary angiography(ICA)within one month after CCTA were retrospectively collected from December 2020 to March 2022 at Changzhou Second People’s Hospital affiliated with Nanjing Medical University.They were diagnosed with CAD based on ICA results showing coronary lumen stenosis greater than 50 percent.There are 85 cases in the CAD group and 72 cases in the non-CAD group.A stratified sampling method was used to randomly divide the 157 patients into a training group(111 patients,60 in the CAD group and 51 in the non-CAD group)and a test group(46 patients,25 in the CAD group and 21 in the non-CAD group)in a ratio of 7:3.The acquired CCTA image data were transferred to the workstation of Coronary Doc,the intelligent analysis system for post-processing.Fat attenuation index(FAI)of the left anterior descending(LAD),left circumflex branch(LCX)and right coronary artery(RCA)were calculated using the FAI automatic calculation function.The differences in pericoronary FAI values between the three vessels were compared,and the final radiomic signature analysis was performed using PCAT of the RCA.All CCTA data were exported from the picture archiving and communication system(PACS)in digital imaging and communications in medicine(DICOM)format and imported into CQK software for CCTA analysis.Coronary segmentation was performed on the images to obtain the RCA,and a region dilation operation was performed on the segmented RCA,with the dilation area equal to the diameter of the vessel.Then the externally expanded region is thresholded and the voxels with gray values from-190 to-30 HU are retained as the Region of Interest(ROI)of PCAT for subsequent extraction of radiomics features.Both max-relevance and min-redundancy(m RMR)and least absolute shrinkage and selection operator(LASSO)regression algorithms were used to filter the radiomics features.Logistic regression analysis was used to identify clinical factors at high risk for coronary heart disease,and models were established by combining radiomics features and clinical risk factors,and normogram was constructed.The efficacy of the normogram was assessed using the receiver operating characteristic(ROC)and calibration curve,and the clinical utility of the normogram was assessed using decision curve analysis(DCA).ResultsFollowing logistic regression analysis,the clinical model was constructed by combining triglyceride(OR 3.03,95%CI 1.53-5.97,P=0.001)and SYNTAX score(OR1.08,95% CI 1.03-1.13,P=0.002),and the AUC in the training and test groups were 0.76 and 0.77,sensitivity were 66.7% and 72.0%,specificity were 76.5% and 66.7%.A total of 11 radiomic features based on RCA pericoronal adipose tissue were selected for building the radiomics model,and their AUCs were 0.73 and 0.70 in the training and test groups,respectively,with sensitivities of 73.3% and 56.0% and specificities of 60.8%and 66.7%.The AUCs of nomogram based on the clinical-radiomics model were 0.82 and 0.81 in the training and test groups,respectively,with sensitivities of 90.0% and 71.0% and specificities of 56.7% and 80.0% for predicting CAD.The calibration curves of the nomogram were close to the ideal line in both the training and test groups.The DCA results showed that the net clinical benefit of the normogram model was better than the clinical model and the radiomics model at threshold probabilities of <53%,60%-78%,and 88%-100% in the training group,and 40%-66% in the test group.ConclusionThe nomogram model constructed by PCAT radiomic features of CCTA combined with clinical indicators TG and SYNTAX score has better predictive value for coronary artery disease than the radiomics model and clinical model alone. |
PDF Full Text Request