Ulcerative Colitis Severity Classification and Localized Extent (UC-SCALE): An Artificial Intelligence Scoring System...

Area of Science:

• Gastroenterology
• Medical Imaging Analysis
• Computational Pathology

Background:

• Current ulcerative colitis (UC) scoring methods like MCES focus on the worst segment, neglecting overall disease extent.
• There is a need for comprehensive, automated tools to evaluate both endoscopic severity and extent in UC.

Purpose of the Study:

• To introduce the Ulcerative Colitis Severity Classification and Localized Extent (UC-SCALE) algorithm for automated UC endoscopic assessment.
• To provide a tool that evaluates both disease severity and extent in UC patients.

Main Methods:

• The UC-SCALE algorithm incorporates a quality filter for colonoscopy videos, an MCES scoring system for each frame, and a camera localization algorithm.
• The system was trained and validated on 4326 sigmoidoscopy videos from Etrolizumab clinical trials.

Main Results:

• UC-SCALE demonstrated high agreement with central and local expert readings (κ = 0.80 and 0.84, respectively).
• The algorithm's scores correlated significantly with clinical markers such as calprotectin, C-reactive protein, patient-reported outcomes, Physician Global Assessment, and Geboes Histologic scores (rs 0.40-0.55, p < 0.0001).
• UC-SCALE effectively assessed changes in endoscopic severity from baseline to induction.

Conclusions:

• The automated UC-SCALE system provides accurate, objective, and localized endoscopic severity assessment for UC.
• UC-SCALE's topological representation of disease severity correlates with clinical metrics, enhancing evaluation in clinical trials and practice.