Abstract
Abstract
Aims
Acute coronary syndrome (ACS) exhibits marked sex-related differences in clinical presentation and underlying pathophysiology. However, whether these differences translate into variations in culprit lesion prediction remains unclear. We investigated the sex-specific performance of conventional coronary computed tomography angiography (CCTA) and artificial intelligence–enabled quantitative coronary plaque and hemodynamic assessment (AI-QCPHA) for ACS culprit detection.
Methods and results
In this sub-study of the EMERALD II trial (NCT03591328), patients with ACS who underwent CCTA 1 month to 3 years prior to the event were analyzed. Culprit and non-culprit lesions were defined by invasive coronary angiography at the time of ACS. Diagnostic performance of obstructive stenosis (CAD-RADS ≥3) and high-risk plaque (HRP) criteria was compared between women and men. Information gain–based feature selection identified sex-specific AI-QCPHA best predictors across five feature clusters, and their incremental predictive value was evaluated using receiver operating characteristics reported as area under the curve (AUC). A total of 351 patients (90 women, 261 men) with 2,451 lesions were included. Conventional CCTA performed similarly across sexes for obstructive stenosis, whereas HRP differed by sex. AI-QCPHA identified the same best features for ACS culprit prediction in both sexes, and integration of these features improved discriminatory performance beyond conventional CCTA (women AUC 0.84 vs. 0.78, p = 0.002; men 0.82 vs. 0.75, p
Conclusion
AI-QCPHA enhances culprit lesion prediction in obstructive ACS beyond conventional CCTA in both sexes, supporting its potential role for improved culprit-lesion discrimination.