Cardiologist user experience of artificial intelligence-based quantitative coronary angiography

Abstract

Background: Artificial intelligence-assisted quantitative coronary angiography (AI-QCA) has been developed to enable the automated, objective assessment of coronary artery stenosis without human intervention. Previous studies have shown its accuracy compared with manual QCA and intravascular ultrasound. In this study, we aimed to evaluate cardiologists&apos; experience of analyzing coronary lesions with AI-QCA. Methods: Ten board-certified cardiologists from multiple centers specializing in coronary intervention, with varying periods of experience, participated in this study. They analyzed angiograms from 180 patients with marked coronary stenosis requiring coronary revascularization. Correlations between manual QCA and AI-QCA were measured by using Pearson&apos;s or Spearman&apos;s correlation coefficients. Results: The average System Usability Scale (SUS) score was 66.7, indicating marginal high acceptability. The angiographic frame selected by the cardiologists with AI-QCA assistance was within five frames of that elected by the QCA analyst in 64.2% of cases. Furthermore, the time taken by cardiologists to analyze angiograms with AI-QCA assistance was 1.5 +/- 0.9 s, significantly lower than that required by an expert analyst to perform manual QCA (88.1 +/- 35.5 s, P<0.001). Key angiographic variables, such as reference vessel diameter (RD), minimal lumen diameter (MLD), diameter stenosis (DS), and lesional length (LL), showed moderate-to-strong correlations between AI-QCA and manual QCA (e.g., distal reference diameter, R=0.74). Conclusions: This prospective study showed that automated analysis with AI-QCA can be performed with an acceptable user experience as well as minimal human intervention and little additional time. Therefore, the application of AI-QCA in the Cath lab is feasible and potentially helpful during coronary angiography (CAG) and intervention.