AI-enhanced coronary CT improves ACS culprit-lesion discrimination in women and men

A sub-study of EMERALD II examined sex-specific performance of conventional coronary computed tomography angiography (CCTA) and AI-enabled quantitative coronary plaque and hemodynamic assessment for identifying subsequent acute coronary syndrome (ACS) culprit lesions. The analysis included 351 patients—90 women and 261 men—with 2,451 lesions. CCTA had been performed 1 month to 3 years before ACS; invasive coronary angiography at the event defined culprit and non-culprit lesions.

Conventional CCTA showed similar performance between sexes for obstructive stenosis, defined as CAD-RADS ≥3, whereas performance of high-risk plaque criteria differed by sex. Feature selection identified the same best AI-derived predictors in women and men. Adding these features improved the area under the receiver operating characteristic curve from 0.78 to 0.84 in women (p=0.002) and from 0.75 to 0.82 in men (p<0.001). Delta FFRCT had the highest discriminatory performance among individual predictors in both groups.

The findings support further evaluation of AI-enhanced CCTA for culprit-lesion discrimination in obstructive ACS. However, this analysis used imaging obtained before events in patients who subsequently developed ACS; it does not establish prospective clinical utility or improved patient outcomes. Women represented a smaller subgroup, and the available abstract does not describe independent validation or provide detailed sex-specific high-risk plaque results.