Identification of the risk of coronary atherosclerosis using opportunistic screening chest CT-derived body composition phenotypes combined with clinical information in middle-aged and older adults

Non-contrast chest computed tomography (CT) scans contain a large amount of background redundant information, but their clinical utilization is currently insufficient. This study aimed to assess whether non-contrast CT-derived body composition phenotypes combined with clinical information could opportunistically identify coronary atherosclerosis (CA) risk in middle-aged and older adults. This retrospective study included 1,196 adults who underwent non-contrast chest CT and were divided into a training set ( n = 837) and a validation set ( n = 359). An external cohort ( n = 53) also served as validation. Body composition phenotypes were automatically extracted from chest CT and selected by least absolute shrinkage and selection operator (LASSO) regression. Clinical variables, including demographics, medical history, and laboratory parameters before imaging, were collected and screened by logistic regression. Models were developed using generalized linear model (GLM), support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost). Performance was evaluated by area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. Among all models, the combined GLM achieved AUCs of 0.817 (95% confidence interval [CI], 0.788–0.845), 0.810 (95% CI, 0.760–0.852), and 0.726 (95% CI, 0.562–0.851) in the internal training, internal validation, and external validation sets, respectively. SHapley Additive exPlanations (SHAP) analysis identified skeletal muscle mean attenuation as the strongest contributor in image model. Features derived from non-contrast chest CT, combined with clinical information, can opportunistically identify CA risk in middle-aged and older adults, extending the utility of chest CT beyond its primary diagnostic purpose.