Cambridge Scientists Create AI Tool That Accurately Diagnosis Celiac Disease
Scientists at the University of Cambridge have created a robust AI algorithm that can diagnose celiac disease from images of biopsies with over 97% accuracy, a significant breakthrough in the application of artificial intelligence to medical diagnosis. The device is intended to help pathologists by recognizing evidence of the autoimmune disease more quickly and accurately, potentially being particularly useful in less well-funded healthcare systems.
The findings were published on March 27 in the New England Journal of Medicine AI.
Celiac disease is diagnosed by analyzing duodenal biopsies for villi damage caused by gluten, but the process can be subjective and inconsistent. The new AI model seeks to eliminate this variability by using machine learning to consistently analyze biopsy images.
AI Model Trained on Diverse Biopsy Data
The AI algorithm was trained on over 3,400 biopsy images from five hospitals with scanners from four manufacturers, ensuring reliable performance across various conditions. In a validation test with 650 unseen biopsy images, the AI accurately identified celiac disease in over 97% of cases, achieving over 95% sensitivity and nearly 98% specificity, demonstrating its high reliability for detection and exclusion.
Outperforming Human Diagnosticians in Consistency
Previous studies conducted by the Cambridge team have revealed that pathologists may disagree on diagnoses in over 20% of cases. In this study, when four experienced pathologists reviewed a set of 30 slides, they were just as likely to agree with the AI’s diagnosis as with another human expert, showing the tool’s clinical potential.
Dr Florian Jaeckle, also from the Department of Pathology, and a Research Fellow at Hughes Hall, Cambridge, said: “This is the first time AI has been shown to diagnose as accurately as an experienced pathologist whether an individual has coeliac or not. Because we trained it on data sets generated under a number of different conditions, we know that it should be able to work in a wide range of settings, where biopsies are processed and imaged differently.”
Also read: China’s PLA Integrates DeepSeek AI Tool for Non-Combat Support Functions
A Scalable Tool with Real-World Impact
The flexible AI tool, built upon a set of biopsy images under diverse imaging and processing conditions, can be used in both high- and low-resource environments. It seeks to accelerate diagnosis, lower costs, and enable pathologists to concentrate on more challenging cases. The second phase is to test the tool in a larger-scale clinical trial to advance towards regulation approval and deployment within the UK's NHS.
Lyzeum Ltd was established by Professor Elizabeth Soilleux and Dr. Jaeckle to take their AI technology to market with the backing of Coeliac UK, Innovate UK, and NIHR, and in partnership with patient groups to drive the acceptance of AI in medical diagnostics.
