Computer Vision News - April 2023

25 IBD Scoring with Clario Clario , a leading healthcare research and technology company that generates the richest clinical evidence for the clinical trials industry and GI Reviewers , LLC Gastroenterology consultants and central readers, together with RSIP Vision , an experienced developer of groundbreaking AI technologies for medical imaging, today announced their new AI-based scoring system for inflammatory bowel diseases. The solution is expected to improve the reproducibility of colonoscopy video scoring, optimize workflow by shortening the time required for a human expert reader to score, and ultimately improve recruitment and reduce costs associated with clinical trials. Peer-reviewed results demonstrating the human-level performance of this innovative system were presented to the pharmaceutical and clinical trials industry community at the prestigious European Crohn’s and Colitis (ECCO) conference this week. Inflammatory bowel diseases (IBD) such as Ulcerative Colitis (UC) and Crohn’s disease are mostly diagnosed by performing a colonoscopy. The level of disease is assessed by visual findings such as ulcers, erosions, and erythema, which are graded using visual scoring approaches. However, IBD scoring is a complex and time-consuming task, and reader assessments are often subjective, based on their training, experience and individual interpretation of the colonoscopy. The new AI-based system, developed at RSIP Vision under the scientific leadership of Clario and the clinical guidance from GI Reviewers, automatically computes the Mayo Endoscopic Score (MES) from colonoscopy videos in Ulcerative Colitis patients. The system was trained on hundreds of videos from multiple sites and devices, containing >100,000 images, and all scored under expert gastroenterologist supervision. Algorithmperformance, as assessed by the Kappa metric, was found to be comparable to the performance of human reviewers. Building on the successes in UC, the system is being extended to support Crohn’s Disease scoring. This technology will allow generation of rich and detailed data from IBD videos, and increase the automation of clinical trial enrollment procedures. “Central reading has become common in gastroenterology clinical trials due to the need for an unbiased, independent review of images ensuring the most reliable outcome data for assessing treatment efficacy,”  said Dr. Marcela Vieira, Medical Director of Gastroenterology at Clario. “Our new AI-based solution will reduce known limitations of central reading, suchas fatigueand inter-reader variability, and improve the reproducibility and speed of the readings. Using AI to support our reviewers is improving our capabilities across a range of clinical trial use cases, and will continue to be a game changer for medical research.” Endoscopicvideos suchascolonoscopiespresent steepchallenges forAI-basedprocessing. Implementing this successful solution requires merging state-of-the-art neural network architectures with strong, robust, and ongoing validation processes, integrating the work of the clinical team as an integral part of the algorithm development.