nnU-Net-based whole-coronary pericoronary adipose tissue radiomics combined with clinical factors for predicting coronary plaque development in individuals with normal coronary arteries

Effective early prediction of coronary artery disease remains challenging. This study aims to investigate whether radiomic features of pericoronary adipose tissue (PCAT) derived from the whole-coronary tree using nnU-Net, combined with PCAT metrics and clinical factors, can predict coronary plaque development during follow-up in individuals with normal baseline coronary arteries. In this retrospective study, 210 patients with normal baseline coronary CT angiography (CCTA) findings who underwent follow-up CCTA were included and classified into plaque-positive and plaque-negative groups. Baseline clinical data and PCAT metrics (fat attenuation index [FAI] and fat volume [FV]) were collected. Radiomic features were extracted from whole-coronary PCAT based on nnU-Net. Patients were randomly divided into training and validation sets (7:3). Five logistic regression models (clinical, PCAT, clinical–PCAT, radiomics, and combined) were constructed and compared using receiver operating characteristic analysis, calibration curves, and decision curve analysis. During follow-up, 77 of 210 patients (36.7%) developed coronary plaques. One clinical factor, three PCAT metrics, and 20 radiomic features were selected. The combined model demonstrated the best performance, with training/validation set AUC values of 0.931 (95% CI 0.889–0.973) and 0.896 (95% CI 0.821–0.970). Calibration and decision curve analyses demonstrated good agreement and clinical utility. An automated nnU-Net-based whole-coronary PCAT radiomics model, integrated with clinical factors, demonstrates potential for early risk stratification of plaque development in individuals with normal coronary arteries.