Colorectal polyp detection in colonoscopy images using YOLO-V8 network
Mehrshad Lalinia, Ali Sahafi
Abstract
Abstract Gastrointestinal tract disorders, including colorectal cancer (CRC), impose a significant health burden in Europe, with rising incidence rates among both young and elderly populations. Timely detection and removal of polyps, the precursors to CRC, are vital for prevention. Conventional colonoscopy, though effective, is prone to human errors. To address this, we propose an artificial intelligence-based polyp detection system using the YOLO-V8 network. We constructed a diverse dataset from multiple publicly available sources and conducted extensive evaluations. YOLO-V8 m demonstrated impressive performance, achieving 95.6% precision, 91.7% recall, and 92.4% F1-score. It outperformed other state-of-the-art models in terms of mean average precision. YOLO-V8 s offered a balance between accuracy and computational efficiency. Our research provides valuable insights into enhancing polyp detection and contributes to the advancement of computer-aided diagnosis for colorectal cancer.