Artificial intelligence-based quantitative coronary angiography (AI-QCA) has recently emerged as a promising tool for real-time lesion assessment in cardiology. We aimed to validate a novel AI-QCA software, trained on a Korean dataset, in a European cohort. We analyzed 556 lesions from 252 subjects in two European datasets. The AI-QCA system performed automated vessel segmentation and measurements of minimum lumen diameter, proximal and distal reference diameters, percent diameter stenosis (%DS) and lesion length. The performance of AI-QCA was assessed using both automated and manual frame selection methods, with all measurements validated against expert manual QCA. AI-QCA achieved a lesion detection rate of 86.2% in automated frame selection. AI-QCA and manual QCA showed strong agreement (Pearson’s r > 0.90, R2 > 0.8 for all QCA measurements). For %DS categorization (<50%, 50% to <70%, and ≥70%), 433 lesions were classified into the same category by both methods, with a weighted κ of 0.832 (95% CI, 0.743–0.922). Vessel segmentation achieved a mean DSC of 0.953. This study validated the performance of AI-QCA using a European dataset and demonstrated high lesion detection rate and its strong agreement with manual QCA, which supports its applicability for real-time clinical decision-making during percutaneous coronary intervention.
Lee et al. (Fri,) studied this question.