The prognostic value of the AI-based Red Density score in ulcerative colitis – the PROCEED-UC trial

作者：P Sinonquel, P Bossuyt, S John, C Rupert, Marietta Iacucci, Uday Nagesh Shivaji, Hiroshi Nakase, Stephanie Van Aelst, S Pillai, Zainab Abdawn, S Sugita, Sara McCartney, Paul Harrow, A Repici, Alessandro Armuzzi, A Dal Buono, R Gabbiadini, F Furfaro, Silvio Danese, Talat Bessissow, TJ Rath, D H Yang, Bram Verstockt, S Vermeire, G De Hertogh, Raf Bisschops · 发表于：Endoscopy · 年份：2026 · DOI：10.1055/s-0046-1820719 · 研究领域：Inflammatory Bowel Disease、Radiomics and Machine Learning in Medical Imaging、Inflammatory Biomarkers in Disease Prognosis

Aims Red Density (RD) is an automated endoscopic scoring tool, and ESGE innovation of the year 2018, developed for the objective evaluation of endoscopic disease activity in ulcerative colitis (UC). RD correlates with established histological and endoscopic scoring systems. The PROCEED-UC trial aimed to assess and validate the predictive value of the Red density score for sustained clinical remission. Methods Patients with UC in clinical remission included in 10 centers in the PROCEED-UC study (NCT04408703) underwent blinded endoscopic assessment with RD in the rectum and sigmoid at baseline, week 52 or early termination (ET) visit. Protocolised biopsies were scored for the Geboes score (GBS), Robart’s Histological Index (RHI), Nancy Histological Index (NHI) and Picasso Histological Remission Index (PHRI) in a blinded fashion after initial scoring convention training. Primary endpoint was sustained clinical remission ((SCCAI)<3 AND no new UC treatment, escalation of therapy, UC related hospitalization or colectomy). Comparison of the means using t-test, cut-off assessment through receiver operating curve analysis (ROC), and correlation testing based on Spearman analysis. Results In total 212 patients from 10 centers were enrolled, 120 completed the 52 weeks follow-up period, 16 had a relapse before 52 weeks, 36 did not complete 52 weeks follow-up and 40 patients had to be excluded (24 due to technical issues and 16 drop outs). A total of 4343 RD images of 1064 colonic segment...