Polypoid Lesion Segmentation Using YOLO-V8 Network in Wireless Video Capsule Endoscopy Images

Abstract

Gastrointestinal (GI) tract disorders are a significant public health issue. They are becoming more common and can cause serious health problems and high healthcare costs. Small bowel tumours (SBTs) and colorectal cancer (CRC) are both becoming more prevalent, especially among younger adults. Early detection and removal of polyps (precursors of malignancy) is essential for prevention. Wireless Capsule Endoscopy (WCE) is a procedure that utilises swallowable camera devices that capture images of the GI tract. Because WCE generates a large number of images, automated polyp segmentation is crucial. This paper reviews computer-aided approaches to polyp detection using WCE imagery and evaluates them using a dataset of labelled anomalies and findings. The study focuses on YOLO-V8, an improved deep learning model, for polyp segmentation and finds that it performs better than existing methods, achieving high precision and recall. The present study underscores the potential of automated detection systems in improving GI polyp identification.

Automatic lesion detection for narrow-band imaging bronchoscopy

“…71 ). Although the size of our dataset is comparable to the datasets used in many other recent endoscopic video analysis studies, [32][33][34][35]38,41,42 further work is needed to better assess the potential of our method. In particular, to assess the method's generalizability across more diverse patient populations and clinical settings, including among physicians at community-based hospitals, a more extensive multicenter study would need to be performed.…”

“…We note that the number of frames and airway exam videos used to develop the detection network is comparable in size to the datasets used in many of the endoscopy studies highlighted in Sec. 1, such as: (1) the EDD2020 dataset used in the EndoCV challenge; 42 (2) the CVC-Colon, CVC-Clinic, ETIS-LaribPolyp, and Kvasir-SEG public colonoscopy databases; [32][33][34][35] and (3) the in-house datasets used by Kanesaka et al 38 and Sahafi et al 41 In general, given the 30 frames per second (fps) video rate during an airway exam, every lesion appeared in >1 frames, with a typical lesion appearing in over 1 or more seconds of video. Thus considerable correlation (i.e., small variations in lesion appearance) typically exists across many of frames constituting the subsequence depicting a given lesion.…”

“…Ali et al 40 integrated the YOLO V3 network (YOLO, you only look once) and several discriminative detection/segmentation models for assessing visual quality of general endoscopic videos. Sahafi et al 41 also adapted a YOLO network to detect polypoid lesions in WCE. In addition, Ali et al 42 reported on the EndoCV challenge results of 17 teams presenting deep learning methods for detecting artifacts and disease instances in gastrointestinal endoscopy video.…”