A survey on contemporary computer-aided tumor, polyp, and ulcer detection methods in wireless capsule endoscopy imaging

Abstract

Wireless capsule endoscopy (WCE) is a process in which a patient swallows a camera-embedded pill-shaped device that passes through the gastrointestinal (GI) tract, captures and transmits images to an external receiver. WCE devices are considered as a replacement of conventional endoscopy methods which are usually painful and distressful for the patients. WCE devices produce over 60,000 images typically during their course of operation inside the GI tract. These images need to be examined by expert physicians who attempt to identify frames that contain inflammation/disease. It can be hectic for a physician to go through such a large number of frames, hence computer-aided detection methods are considered an efficient alternative. Various anomalies can take place in the GI tract of a human being but the most important and common ones and the aim of this survey are ulcers, polyps, and tumors. In this paper, we have presented a survey of contemporary computer-aided detection methods that take WCE images as input and classify those images in a diseased/abnormal or disease-free/normal image. We have considered methods that detect tumors, polyps and ulcers, as these three diseases lie in the same category. Furthermore, general abnormalities and bleeding inside the GI tract may be the symptoms of these diseases; so an attempt is also made to enlighten the research work done for abnormalities and bleeding detection inside WCE images. Several studies have been included with in-depth detail of their methodologies, findings, and conclusions. Also, we have attempted to classify these methods based on their technical aspects. A formal discussion and comparison of recent review articles are also provided to have a benchmark for the presented survey mentioning their limitations. This paper also includes a proposed classification approach where a cascade approach of neural networks is presented for the classification of tumor, polyp, and ulcer jointly along with data set specifications and results. (C) 2020 Elsevier Ltd. All rights reserved.