2021/09/29 by Evan Schwab, Gabriela Oana Cula, Schwab, Evan +11 · 3 citations
Medicine · #Artificial Intelligence (cs.AI) #Colorectal Cancer Screening and Detection #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Helicobacter pylori-related gastroenterology studies #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Mycobacterium research and diagnosis #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2109.14685
openalex publication_date 2021/09/29 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Ulcerative colitis (UC) is a chronic inflammatory bowel disease characterized\nby relapsing inflammation of the large intestine. The severity of UC is often\nrepresented by the Mayo Endoscopic Subscore (MES) which quantifies mucosal\ndisease activity from endoscopy videos. In clinical trials, an endoscopy video\nis assigned an MES based upon the most severe disease activity observed in the\nvideo. For this reason, severe inflammation spread throughout the colon will\nreceive the same MES as an otherwise healthy colon with severe inflammation\nrestricted to a small, localized segment. Therefore, the extent of disease\nactivity throughout the large intestine, and overall response to treatment, may\nnot be completely captured by the MES. In this work, we aim to automatically\nestimate UC severity for each frame in an endoscopy video to provide a higher\nresolution assessment of disease activity throughout the colon. Because\nannotating severity at the frame-level is expensive, labor-intensive, and\nhighly subjective, we propose a novel weakly supervised, ordinal classification\nmethod to estimate frame severity from video MES labels alone. Using clinical\ntrial data, we first achieved 0.92 and 0.90 AUC for predicting mucosal healing\nand remission of UC, respectively. Then, for severity estimation, we\ndemonstrate that our models achieve substantial Cohen's Kappa agreement with\nground truth MES labels, comparable to the inter-rater agreement of expert\nclinicians. These findings indicate that our framework could serve as a\nfoundation for novel clinical endpoints, based on a more localized scoring\nsystem, to better evaluate UC drug efficacy in clinical trials.\n