Aims

The relationship between the extent and composition of coronary atherosclerosis and the severity of myocardial ischaemia remains incompletely understood. We assessed whether artificial intelligence-guided coronary computed tomography angiography–derived plaque burden and composition correlate with ischaemia severity.
Methods and results

We included 837 symptomatic patients undergoing coronary computed tomography angiography and subsequent 15O-water positron emission tomography myocardial perfusion imaging. Artificial intelligence–guided coronary computed tomography angiography was used to quantify plaque features—diameter stenosis, percent atheroma volume (PAV), percent non-calcified plaque volume (NCPV), and percent calcified plaque volume (CPV)—per patient and per major coronary artery (LAD, LCx, RCA). Ischaemia severity was classified into four categories based on regional hyperaemic myocardial blood flow. Increasing severity of ischaemia was associated with higher diameter stenosis and plaque burden (PAV, NCPV, CPV) on patient level and in all major coronary territories (overall P < 0.001). The LAD consistently demonstrated higher atherosclerotic burden as compared to the LCx and RCA. Ordinal logistic regression confirmed that diameter stenosis (OR 1.02–1.03, P < 0.001) and NCPV (OR 1.04–1.05, P = 0.011–0.031) were significant predictors of ischaemia severity in all coronary arteries, while CPV was predictive only in the LAD and RCA (OR 1.03–1.04, P = 0.002–0.015).
Conclusion

Artificial intelligence–guided coronary computed tomography angiography–derived measures of plaque burden and stenosis are associated with the severity of myocardial ischaemia, although overlapping distributions across ischaemia severity indicate that anatomical imaging alone may be insufficient for accurate phenotyping of flow-limiting CAD. These findings encou