Executive Summary : | The burden of Gastro-Intestinal (GI) and liver diseases in Asian countries like India, China, and Japan has increased due to various environmental factors, including industrialization, changes in nutrition, and increased antibiotic use. Small Bowel Capsule Endoscopy (SBCE) is a technique used for diagnosing, treating, and managing these diseases, but it faces challenges such as cost and evaluation time without affecting diagnostic precision. Manual interpretation and false-positive rates of abnormalities in CE videos are high due to extensive analysis and impairment of mucosal frames. Artificial intelligence (AI) is predicted to significantly impact CE technology in abnormality detection, quality assessment, cleansing score system, video summarization, and artefact removal. To assess the reliability of computer-aided findings in SBCE, a high-quality, multi-label, and medically validated data development is needed, and it's cleansing during patient preparation and post-processing of obtained videos is crucial. Currently, a standardized, computer-operated SB cleansing scale is not available. The proposed project aims to deliver a fully automatic, explainable, multiple abnormality identification framework with reduced reading time and effective visual quality assessment for SBCE. A high-quality, multi-label, medically validated, large Indian population-based SBCE AI database will be developed with the help of a team of doctors from AIIMS Delhi, followed by the framework development for multiple abnormality identification and associated cleansing score system. |