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

Area of Science:

• Gastroenterology and Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Gastrointestinal (GI) tract disorders, including small bowel tumours (SBTs) and colorectal cancer (CRC), are increasing, particularly in younger populations.
• Early detection and removal of polyps, precursors to malignancy, are critical for preventing GI cancers.
• Wireless Capsule Endoscopy (WCE) generates vast amounts of imaging data, necessitating efficient automated analysis for polyp detection.

Purpose of the Study:

• To review and evaluate computer-aided polyp detection methods for WCE imagery.
• To assess the performance of the YOLO-V8 deep learning model for automated polyp segmentation in WCE images.

Main Methods:

• A review of computer-aided detection approaches for polyps in GI tract imaging.
• Evaluation of the YOLO-V8 deep learning model using a labelled dataset of GI anomalies and findings from WCE.
• Comparative analysis against existing polyp detection methods.

Main Results:

• The YOLO-V8 deep learning model demonstrated superior performance in polyp segmentation compared to existing methods.
• High precision and recall were achieved by the YOLO-V8 model in identifying polyps from WCE images.
• The study validates the effectiveness of automated systems for enhancing GI polyp identification.

Conclusions:

• Automated polyp segmentation using deep learning, specifically YOLO-V8, holds significant potential for improving the early detection of GI polyps.
• This technology can aid in the prevention of gastrointestinal cancers by facilitating timely polyp removal.
• The findings highlight the value of advanced AI in managing public health issues related to GI disorders.