| Objective:1.To evaluate the clinical application value of constructing a model based on coronary computed tomographic angiography(CCTA)pericoronary fat attenuation index(FAI)to predict the occurrence of acute coronary syndrome(ACS).2.To evaluate the clinical application value of constructing a model based on CCTA pericoronal fat imaging radiomics to predict the occurrence of ACS.Materials and Methods:The imaging data and clinical data(including gender,age,BMI,traditional cardiovascular disease risk factors,etc)of 171 patients who underwent CCTA examination at the Hospital of Changzhou,Nanjing Medical University from January2018 to March 2022 were retrospectively analyzed,and those who were clinically diagnosed with ACS within two years were included in the ACS group(91patients),including 66 cases of unstable angina,8 cases of ST elevation myocardial infarction and 17 cases of non-ST elevation myocardial infarction;patients matched according to age and gender who had CCTA at our hospital during the same period and had not had ACS within two years were included in the n ACS group(80patients).Subsequently,all patients were randomly divided into training group(64 ACS group and 56 n ACS group)and test group(27 ACS group and 24 n ACS group)according to a ratio of 7:3.In the post-processing workstation,the coronary arteries were reconstructed and post-processed by techniques such as VR,MIP,CPR and MPR,and the reconstructed CCTA images were imported into CQK software for region of interest(ROI)delineation of the adipose tissue around the coronary arteries,and the FAI values of the proximal right coronary artery(RCA)(starting at 10 mm from the ostium and 40 mm in length)were automatically extracted from all patients,and the FAI values of patients in the ACS group and patients in the n ACS group were compared.The ROI was extracted for fat-image omics characteristics using the radiomics software package,and the extracted characteristics were screened and dimensionality reduced using maximum correlation minimum redundancy(m RMR)and minimum absolute value convergence and selection operator(LASSO)linear regression to obtain the optimal radiomics characteristics for predicting ACS to construct the image radiomics score(Rad-score).Then FAI model,radiomics model and prediction model were constructed based on FAI and Rad-score,and receiver operating characteristic curve(ROC)and area under the curve(AUC)were used to assess the diagnostic efficacy of different models,and the accuracy,sensitivity and specificity of the prediction model were calculated;and decision curve and calibration curve analysis(DCA)were plotted to assess the accuracy and clinical use of the model.Results:Patients in the ACS group had higher FAI values measured in the right coronary artery(RCA)(-76.38 ± 7.7 HU)than patients in the n ACS group(-82.06 ± 7.53 HU),and the difference was statistically significant(P value < 0.001).A total of 1158image-omics features were extracted from the pericoronal fat ROIs delineated by CCTA examination,and a total of 14 optimal omics features were selected to establish the Rad-score based on m RMR and LASSO linear regression.The area under the curve(AUC)of FAI model,imaging omics model and comprehensive model for predicting ACS in the training group was 0.738,0.747 and 0.849,respectively,with accuracy of74.2%,70.0% and 83.3%,sensitivity of 82.8%,67.2% and 89.1%,and specificity of64.3%,83.3% and 76.8%.The AUCs of the three models for predicting ACS in the test group were 0.650,0.739,and 0.813;the accuracies were 64.7%,74.5%,and 84.3%;the sensitivities were 44.4%,96.3%,and 85.2%;and the specificities were 87.5%,50.0%,and 83.3%,respectively.Conclusion:The predictive model based on CCTA pericoronal fat can provide some information for predicting the occurrence of ACS,while the diagnostic efficacy of the model based on image-omics is better than that based on pericoronal FAI,and the diagnostic efficacy of the model constructed by the combination of the two is the best. |
PDF Full Text Request