Detection of Pathological Markers in Colonoscopy Images using YOLOv7

Colonoscopy is a procedure for inspecting the colon of a patient. On colonoscopy images, anatomical landmarks, irritated and swollen tissue, ulcers, and polyps in the colorectal region can all be detected. These images are critical in identifying a variety of colorectal pathological disorders. The third leading cause of cancer-related death in both men and women is colorectal cancer. Many colorectal cancers can be avoided with routine screening. Polyp identification during colonoscopy may be prone to human error, such as missed polyps, which can lead to a missed diagnosis of colorectal cancer. In this study, we employ YOLOv7, enabling real-time and more accurate polyp detection that can be used in real-world colonoscopy. This will be useful for current physicians as well as training new clinicians. With an F1 score of 0.9304, the proposed model is capable of identifying polyps. Our approach outperforms existing methodologies by 7.5 percentage in terms of precision and 1.05 percentage in terms of F1 score.