Abstract 4142296: Radiomic Texture Analysis of Peri-Coronary Adipose Tissue Significantly Improves Risk Prediction of Major Cardiac Events in the CORE320 Study

Introduction: Peri-coronary adipose tissue (PCAT) assessment using coronary computed tomography angiography (CCTA) is an evolving method for risk stratification in non-invasive evaluation of coronary artery disease. Radiomic texture analysis of PCAT has greater predictive value than traditional risk stratification metrics, detecting structural changes linked to heart disease Aim: To assess the effectiveness of radiomic texture features from peri-coronary adipose tissue in predicting cardiac events compared to traditional methods Methods: The CORE320 study enrolled 381 patients referred for invasive coronary angiography (ICA) due to known or suspected coronary artery disease (CAD). Participants underwent CCTA and were monitored for 5 years for major adverse cardiac events (MACE), defined as revascularization, all-cause death, and myocardial infarction. The analysis focused on the proximal 10-50 mm of the right coronary artery (RCA), where 284 RCA vessels were manually contoured using 3D Slicer. Using PyRadiomics Python, we extracted 113 radiomic features from the peri-coronary adipose tissue (PCAT), including previously published PCAT values. Feature importance for predicting a 5-year incidence of major adverse cardiac events (MACE) was determined using the XGBoost algorithm. These features were assessed in a multivariable model using Cox proportional hazards regression on a held-out test set Results: In the complete analysis, MACE occurred in 99 of the 381 patients (28%) during the follow-up period. Traditional measures of PCAT attenuation (HR 0.96, 95% CI 0.75-1.22, p=0.59) did not provide significant predictive value for MACE events. Among the 113 radiomic features extracted from the peri-coronary adipose tissue, the XGBoost classifier identified texture features associated with 5-year MACE after adjusting for clinical risk factors. The top-5 included homogeneity, entropy, energy, correlation, and cluster shade. In multivariate analysis, entropy and homogeneity remained independently associated with MACE after adjustment for clinical risk factors. The addition of these two radiomic features to a model with clinical risk factors significantly improved discrimination for 5-year MACE (C-statistic improved from 0.69 to 0.74, p=0.003) Conclusions: Radiomic texture analysis of PCAT on CCTA can significantly enhance risk stratification for major adverse cardiac events in patients with suspected coronary artery disease, surpassing traditional clinical risk factors