From Colon to Uterus: Potential of YOLOv7 for Real-Time Polyp Detection in Hysteroscopy

Polyp detection is critical in both gastrointestinal and gynecological practice, with early identification significantly reducing cancer risk. While deep learning models like YOLOv7 have demonstrated high ac-curacy in colonoscopic polyp detection, their application in hysteroscopy remains unexplored. This study systematically reviews YOLOv7’s performance in colonoscopy and assesses its feasibility for hys-teroscopic polyp detection through a two-phase methodology: (1) a literature review of YOLOv7 applica-tions in polyp detection, and (2) a comparative analysis of technical transferability to hysteroscopy. Find-ings reveal that YOLOv7 achieves precision up to 98.8%, recall up to 96.3%, and real-time processing speeds (~20 ms/frame) in colonoscopy, often enhanced by attention mechanisms and advanced loss functions. However, no studies directly evaluate YOLOv7 for hysteroscopy, highlighting a critical re-search gap. Technical parallels between colonoscopic and hysteroscopic imaging—such as lesion mor-phology, illumination challenges, and real-time requirements—suggest plausible adaptability, but do-main-specific validation is lacking. Key limitations include the absence of hysteroscopic datasets and benchmarking. This study underscores the need for dedicated research to bridge this gap, proposing YOLOv7 as a promising candidate for advancing real-time polyp detection in gynecological endoscopy.