P0278Predicting histological activity in Ulcerative Colitis using AI-driven endoscopic disease scoring

Abstract Background Ulcerative colitis trials commonly pair endoscopic and histological assessments to define disease activity. Because endoscopic remission often predicts histological remission, endoscopybased AI that scores disease severity can also infer histological state. Demonstrating this capability would validate model robustness and enable biopsysparing endpoints for clinical trials. Our aim is to evaluate how AI-derived endoscopic Mayo Endoscopic Subscore (MES) and Ulcerative Colitis Endoscopic Index of Severity (UCEIS) metrics predict histological Nancy Index (NI), Robarts Histopathology Index (RHI) and Simplified Geboes Score (SGS) pathologist scores. Methods Endoscopy videos from 55 patients with UC from the WHITS study (NCT05000242) were analysed using a deep learning model pre-trained for MES and UCEIS scoring. 153 endoscopy videos of 20 seconds spanning the point of biopsy were assessed by the AI model. The biopsies were read by two expert pathologists with more than 14 years of experience.AI endoscopic evaluation of active and remission states (MES 1 and UCEIS ≤ 1 and histological remission as SGS ≤ 2A, NI = 0, RHI ≤ 3. Diagnostic accuracy for identifying remission using the AI-assisted endoscopy algorithm was compared with biopsy as the ground truth using AUC; categorical agreement between tests was assessed using quadratic weighted kappa statistics (QWK). Patient demographics: mean age 43 years, 56% male, with varied treatments. Results AI endoscopic assessment of active and remission states showed substantial agreement with pathologist evaluation. Agreement between AI-predicted endoscopic remission and histological remission for MES: QWK 0.66 (SGS), 0.64 (NI) and 0.68 (RHI). For UCEIS, agreement was QWK 0.71 (SGS), 0.67 (NI) and 0.74 (RHI). Accuracy was also high with MES AUCs of SGS 0.84 (95 % CI 0.78–0.89), NI 0.84 (0.79–0.89) and RHI 0.87 (0.82–0.91). For UCEIS AUCs were SGS 0.85 (0.79-0.91), NI 0.84 (0.78-0.90) and RHI 0.87 (0.82-0.92). Conclusion We have developed a tool for objectively scoring endoscopy videos that agrees with histological findings. This demonstrates the model’s capability to accurately assess disease activity in ulcerative colitis in both MES and UCEIS. This provides opportunity for increasing the efficiency and accuracy of clinical trials as well as providing new insight in treatment effects. Furthermore, its ability to predict histological outcomes could help reduce the histological workload in trials. Conflict of interest: Mr. Bogush, Alexander: I disclose the following financial relationship(s) with a commercial interest: Perspectum Ltd Windell, Dylan: I disclose the following financial relationship(s) with a commercial interest​: Perspectum Ltd.​ Ralli, George: No conflict of interest Suzuki, Noriko: No conflict of interest Toskas, Alexandros: No conflict of interest Fryer, Eve: None Wakefield, Phil: No conflict of interest Goldin, Rob: None Landy, Jonathan: No conflict of interest