Real-Time Detection and Classification of Stenosis in Coronary Arteries: An AI-Driven Approach

Abstract

Coronary artery disease (CAD) is one among world's major causes of morbidity and mortality, so early and precise detection becomes imperative to improve patient outcome. The current work puts forward real-time stenosis recognition and classification from angiographic images with the help of a YOLOv8 object detector architecture. The five variants, including the YOLOv8n, YOLOv8s, YOLOv8m, YOLOv8l, and YOLOv8x, were compared with important performance metrics. In them, YOLOv8s demonstrated speed-accuracy tradeoff, achieving 78.13 FPS speed, 0.967 precision, and 0.981 Map 50. The system was verified with angiographic images and proved to dynamically process each frame, with stenotic regions properly identified and classified real time. Comparative study with existing detecting models guaranteed proposed approach achieved higher speed and diagnostic capability. The findings justify CAD real-time diagnostic feasibility with YOLOv8s, with a promising tool to refine precision, reduce human error, and permit timely action within a procedure of coronary angiography. While this study’s application of YOLOv8 for detecting and classifying coronary artery stenosis has demonstrated promising results.