Integrating coronary atherosclerotic plaque burden for enhanced grading of AI-guided CCTA-derived ischemia

Abstract Background The extent and severity of myocardial perfusion abnormalities and absolute myocardial blood flow (MBF) are important for risk stratification in chronic coronary artery disease (CAD). The artificial intelligence -guided quantitative computed tomography ischemia algorithm (AI-QCTischemia) is a novel method for predicting whether myocardial ischemia is likely or unlikely, using data from coronary computed tomography angiography (CCTA). Purpose We hypothesized that incorporating quantitative measures of coronary atherosclerotic plaque burden could enable further risk stratification among patients with abnormal AI-QCTischemia results (i.e., ischemia likely). Methods From an institutional registry, we identified 662 symptomatic patients who underwent [15O]H2O positron emission tomography (PET) stress myocardial perfusion imaging due to suspected obstructive (≥50%) stenosis on CCTA. The CCTA scans were analyzed in blinded fashion using the AI-QCTischemia algorithm, and percent atheroma volume (PAV) was calculated as a per-patient measure of coronary atherosclerotic plaque burden. Patients were followed for the composite endpoint of all-cause mortality and myocardial infarction (MI). Results Among the 662 patients (58% male, age 66 ± 8 years), 416 (63%) patients were classified as abnormal by AI-QCTischemia (i.e., ischemia likely) with PET-measured regional stress MBF of median 1.8 ml/g/min (25th–75th percentiles: 1.3–2.6 ml/g/min). The latter were further categorized into three groups based on PAV tertiles (22%). This classification – combining AI-QCTischemia status with PAV tertiles – was associated with progressively lower regional stress MBF as measured by PET (p