AI-enabled Quantitative High-Risk Plaque Attributes for Predicting Coronary Events in Nonculprit Vessels.

Abstract

Purpose To assess the prognostic value of artificial intelligence (AI)-derived high-risk features obtained from coronary CT angiography (CCTA) in nonculprit vessels of patients who have undergone percutaneous coronary intervention. Materials and Methods This retrospective study included patients who underwent CCTA at a tertiary hospital between June 2013 and June 2023 followed by percutaneous coronary intervention within 3 months. AI-derived high-risk CCTA features were evaluated, including significant stenosis, high-risk plaque, high plaque volume, low CT fractional flow reserve, and high pericoronary adipose tissue attenuation. The primary end point was nonculprit vessel-related major adverse cardiac events (MACE). The prognostic value of high-risk CCTA features was assessed using multivariable Cox regression analyses. Results A total of 1495 patients (mean age, 66 years ± 10; 1100 male patients) with 2014 nonculprit vessels were analyzed with a median follow-up of 3.3 years. MACE occurred in 100 vessels (5.0%). In a multivariable Cox analysis adjusted for high-risk features, vessel-specific CT fractional flow reserve (adjusted hazard ratio, 0.14; 95% CI: 0.03, 0.76; P = .02) and necrotic core volume (adjusted hazard ratio, 1.30; 95% CI: 1.06, 1.59; P = .01) were independent predictors of MACE and showed an incremental prognostic value when added to clinical risk factors (area under the receiver operating characteristic curve, 0.60 vs 0.67; P < .001). Conclusion AI-derived high-risk features at CCTA provided independent and incremental prognostic values for nonculprit vessel-related MACE. Keywords: Computed Tomographic Angiography, High-Risk Plaque, Percutaneous Coronary Intervention Supplemental material is available for this article. © RSNA, 2026.