The use of Capsule Endoscopic Examination Videos in the Detection of Abnormalities in the Gastrointestinal Tract

Abstract

A gastric abnormality involves the stomach and other nearby organs that are involved in digestion. Diagnosing and screening for gastric abnormalities can be time-consuming and challenging as many stomach and digestive disorders have similar symptoms. 60 to 70 million Americans suffer from gastric abnormalities which lead to nearly 250,000 deaths per year according to the ‘National Institute of Diabetes and Digestive and Kidney Diseases’. To overcome the current limitations, our approach uses deep learning (DL) integrated with wireless capsule endoscopy for segmentation of video capsule endoscopy images to detect four different abnormalities in the Gastrointestinal Tract (polyps, Angiectasias, erythema and Lymphangiectasia) and to develop lightweight, low-latency models that can be integrated with low-end endoscopic hardware devices [1]. Deep learning (DL) is a subfield of Machine Learning (ML) that utilizes layered structure of algorithms inspired by the biological neural network of the human brain. DL can reinforce disease diagnosis, interventions; and documenting procedure findings and quality measures [2]. DL has the potential to revolutionize gastrointestinal endoscopy as it can enhance clinical performance and support assessing lesions more accurately when trained by domain experts [3]. Experiments has been conducted on large benchmark dataset of Kvasir-Capsule dataset achieving high segmentation accuracy, sensitivity and specificity of 98.60%, 100% and 76.99%, respectively. The experimental findings demonstrate that the proposed model has achieved enhanced performance in terms of a trade-off between model complexity, metric performances and model parameters.