Automated Image Analysis for Gastrointestinal Polyp and Ulcer Identification towards Sustainable Health Diagnostics
Hasini Dilani Ranasinghe
Abstract
Wireless capsule endoscopy (WCE) is an advanced imaging technology for the diagnosis of patients remotely during gastrointestinal (GI) procedures, thereby improving patient comfort and data resolution. The WCE procedure involves a swallowable miniature camera device equipped with light emitting diodes (LEDs) to record images of the GI tract. These images are then transmitted to gastroenterologists to examine the images to identify clinical conditions or abnormalities such as polyps, lesions, or bleeding, thereby facilitating diagnostic evaluations. The recorded images have a significant amount of redundancy, along with low-resolution or unclear features which need to be removed to extract useful information from the recorded images. A number of deep learning models have been developed for automated detection of polyps and ulcers, each having their benefits and drawbacks. Colorectal polyps exhibit diverse shape, texture, and color features even within a single patient's video, complicating the task of recognizing polyps and ulcers. Here we review the deep learning models for the detection and segmentation of polyps and ulcers from WCE recorded videos, the challenges involved in data segmentation and image processing, and future outlook in automated polyp and ulcer detection.
