Nature Cardiovascular Research Biochemistry, Genetics and Molecular Biology (miscellaneous)
A multi-view video-text foundation model for invasive coronary angiography interpretation and risk prediction
Abstract
Invasive coronary angiography remains central to the management of coronary artery disease, but it is interpreted visually, one projection at a time. Now, in a non-randomized two-center study, a video-text foundation model generates reports that predicts clinical outcomes and aligns more closely with expert assessments than routine single-reader reports.