AI-driven Narrow-band Imaging Score for Disease Assessment and Outcome Prediction in Ulcerative Colitis

Ulcerative Colitis (UC)

Ulcerative Colitis, Virtual Chromoendoscopy, Artificial Intelligence, Disease Assessment, Outcome Prediction

This international multicentre prospective study aims to develop a new simple score using enhanced endoscopic techniques which focus on the vascular features of the colon and reliably distinguish between a quiescent and a mild inflammation in ulcerative colitis (UC). The diagnostic performance of the new score in defining disease activity/remission compared to existing endoscopic and histological scores and predict long-term clinical outcomes will be evaluated. The study also aims to adapt current artificial intelligence (AI) algorithms for enhanced endoscopic techniques to improve standardization in UC disease assessment and outcome prediction.

This is a multicentre prospective international study. This study aims at developing a new simple endoscopic score using white light endoscopy - high definition (WLE-HD), Texture and colour enhancement imaging (TXI), red dichromatic imaging (RDI) and narrow-band imaging (NBI) modes, focusing on vascular features to distinguish between quiescent versus patchy mild Ulcerative Colitis. It will evaluate the new score's diagnostic performance in defining disease activity/remission compared to existing endoscopic and histological scores and predict long-term clinical outcomes. Finally, it also aims to develop and adapt existing artificial intelligence (AI) algorithms according to WLE-HD, TXI, RDI and NBI to grade and standardize endoscopic and histological disease assessment and predict long-term clinical outcomes. The study will be divided in several phases: * In the first phase, the score will be developed on the first 30 consecutive virtual electronic chromoendoscopy (VCE) videos (using TXI-RDI and NBI) of UC patients, with different grade of disease activity. Experts in inflammatory bowel disease (IBD) endoscopy will review images and videos from recruited patients to define the endoscopic mucosal and vascular features of the new score. These will be used for a stepwise discussion. A round table discussion using modified Delphi method will be conducted by experts worldwide to ensure equal participation and identify the best component descriptors of endoscopic vascular healing. The components that will achieve 100% consensus will be selected, and the most important endoscopy predictive variables will be confirmed by using a machine learning technique. Finally, a new endoscopic score will be generated. This should be reproducible, valid and responsive. * In the second phase, the new endoscopic scoring system will be validated in a large cohort of UC patients, focusing on patients with quiescent disease versus patchy mild colitis. Diagnostic accuracy, interobserver agreement and ability to predict clinical outcome according to the new endoscopic score focused on vascular features assessed with VCE will be evaluated * In the third phase, the reproducibility of the new endoscopic scoring system will be evaluated among gastroenterologists with different levels of experience through a short survey and a computerised training module. * In the fourth phase, new and existing AI algorithms will be developed and adapted to these endoscopic videos and histological images to grade and standardize endoscopic and histological disease assessment and predict long-term clinical outcome in UC.

Belgium, Germany, Ireland, Italy, Japan, Singapore