A combined radiomics-clinical model for early detection of coronary plaque development in coronary CT angiography-negative populations

Background: Despite negative coronary computed tomography angiography (CCTA) findings, many patients remain at risk for subclinical atherosclerosis and future cardiovascular events. Our aim was to develop an interpretable combined model integrating pericoronary adipose tissue (PCAT) radiomics features with clinical risk factors to predict newly developed coronary plaques in patients with initially normal CCTA results. Methods: This retrospective study included 947 patients who underwent two CCTA examinations and had normal findings on the initial scan. Based on follow-up CCTA results, patients were classified into new plaque and no new plaque groups. A total of 279 radiomic features were extracted from PCAT on baseline images. Three predictive models were built: a clinical model, a radiomics model, and a combined model. Model performance was evaluated using AUC, calibration, and decision curve analysis. DeLong's test was used to compare model performance. Model interpretability was assessed using Shapley Additive Explanations (SHAP). Results: The combined model outperformed the clinical and radiomics models in both training and validation cohorts. It achieved AUCs of 0.99 (95 % CI: 0.98–0.99) and 0.93 (95 % CI: 0.89–0.97), with sensitivities of 0.92 and 0.86, and specificities of 0.97 and 0.91, respectively. DeLong's test confirmed the superiority of the combined model (P < 0.05). Calibration curves and decision analysis demonstrated excellent consistency and greater clinical utility. Conclusions: The interpretable combined model showed superior performance in predicting early coronary plaque development in CCTA-negative patients. By integrating radiomic and clinical data, it offers a promising tool for risk stratification, enabling earlier intervention and personalized preventive strategies in coronary artery disease management.