Artificial Intelligence Enhances Colonoscopy Evaluation for Crohn's Disease

A groundbreaking study has demonstrated that artificial intelligence (AI) can significantly improve the assessment of Crohn's disease during colonoscopy procedures. Published in the journal Clinical Gastroenterology and Hepatology, the research reveals that AI models can analyze endoscopic images with accuracy comparable to, or even surpassing, experienced gastroenterologists.

In this study, researchers utilized a computer vision AI system trained on 4,487 still images extracted from endoscopic videos of Crohn's disease patients. The AI was tasked with identifying mucosal ulceration, a key indicator of disease severity. Its performance was evaluated against traditional scoring systems, notably the Simple Endoscopic Score for Crohn's Disease (SES-CD), and the assessments made by gastroenterologists.

Results showed that the AI model achieved a higher DICE similarity score (.591) in identifying ulcers compared to the concordance between two physicians (.462). Additionally, the AI's evaluations correlated strongly with SES-CD scores, which are used to quantify ulceration extent but are limited by observer variability and difficulty capturing certain lesion features.

Dr. Ryan W. Stidham of the University of Michigan commented on the implications, highlighting that clinicians often rely on a gestalt judgment of disease severity, which can be subjective. AI-based image analysis offers a more objective and reproducible tool for disease assessment, providing precise measurements during colonoscopy that can improve diagnostic accuracy.

The research involved expert annotations of ulcer areas in endoscopic images, with the AI model reviewing the same set, demonstrating the potential to streamline and standardize Crohn's disease evaluation. The incorporation of AI could be particularly beneficial in areas lacking IBD specialists, assisting less experienced providers and guiding treatment decisions.

Looking ahead, the researchers emphasize that this is an initial step toward more quantifiable and comprehensive disease assessment. Future applications may include aiding in drug development, improving disease monitoring, and providing objective data for clinical decision-making. As Dr. Stidham pointed out, these advancements pave the way for collaborative AI and expert efforts to enhance patient care.

Source: https://medicalxpress.com/news/2025-08-ai-analysis-colonoscopy-crohn-disease.html