Predictive Value of CCTA-based Pericoronary Adipose Tissue Imaging for Major Adverse Cardiovascular Events

To evaluate the ability of the radiomic characteristics of pericoronary adipose tissue (PCAT) as determined by coronary computed tomography angiography (CCTA) to predict the likelihood of major adverse cardiovascular events (MACEs) within the next five years.

In this retrospective casecontrol study, the case group consisted of 210 patients with coronary artery disease (CAD) who developed MACEs within five years, and the control group consisted of 210 CAD patients without MACEs who were matched with the case group patients according to baseline characteristics. Both groups were divided into training and testing cohorts at an 8:2 ratio. After data standardization and the exclusion of features with Pearson correlation coefficients of |r| ≥ 0.9, independent logistic regression models were constructed using selected radiomics features of the proximal PCAT of the left anterior descending (LAD) artery, left circumflex (LCX) artery, and right coronary artery (RCA) via least absolute shrinkage and selection operator (LASSO) techniques. An integrated PCAT radiomics model including all three coronary arteries was also developed. Five models, including individual PCAT radiomics models for the LAD artery, LCX artery, and RCA; an integrated radiomics model; and a fat attenuation index (FAI) model, were assessed for diagnostic accuracy via receiver operating characteristic (ROC) curves, calibration curves, and decision curves.

Compared with the FAI model (AUC=0.564 in training, 0.518 in testing), the integrated radiomics model demonstrated superior diagnostic performance (area under the curve [AUC]=0.923 in training, 0.871 in testing). The AUC values of the integrated model were greater than those of the individual coronary radiomics models, with all the models showing goodness of fit (P > 0.05). The decision curves indicated greater clinical utility of the radiomics models than the FAI model.

PCAT radiomics models derived from CCTA data are highly valuable for predicting future MACE risk and significantly outperform the FAI model.