Coeliac disease is a chronic autoimmune disorder that affects the small intestine and is triggered by the ingestion of gluten in genetically predisposed individuals. In coeliac disease, the consumption of gluten (a protein found in wheat, barley, and rye) elicits an autoimmune reaction that damages the small intestine. This can lead to a wide range of symptoms, including nutrient malabsorption, chronic diarrhoea, unintended weight loss, fatigue, anaemia, dermatitis herpetiformis (a characteristic itchy skin rash), and even infertility.
The clinical presentation can vary significantly from one individual to another, which complicates timely and accurate diagnosis. Notably, even individuals without apparent symptoms (asymptomatic patients) remain at risk for serious long-term complications.
Diagnosis often relies on microscopic analysis of intestinal tissue samples, but the interpretation of these histological images can vary considerably between pathologists. This variability can lead to uncertainty or errors in the diagnostic process.
Recently, to address this challenge, a team of Researchers (Jaekle, F et al. NEJM AI. 2025) turned to artificial intelligence. Specifically, they developed a system based on deep learning, an advanced form of machine learning that enables computers to recognize complex patterns within large volumes of data. In this case, the data consisted of digital images of intestinal tissues.
Using thousands of histological images, the researchers trained an algorithm capable of distinguishing between healthy tissue and tissue affected by coeliac disease. The results were striking: the model achieved a diagnostic accuracy of 94.5%, with a sensitivity of 95% (its ability to correctly identify cases of the disease) and a specificity of 94% (its ability to correctly exclude non-diseased cases). In essence, the AI system performed on par with experienced pathologists.
One of the most promising aspects of this work is the model’s ability to generalize, meaning it performed well even when tested on images from clinical centers different from the one where it was originally trained. In some cases, the algorithm even detected signs of the disease that had been overlooked by human pathologists, suggesting that it could become a valuable assistant in everyday diagnostic work.
This study represents a significant step toward integrating artificial intelligence into clinical practice. While such systems are not intended to replace the clinical judgment of specialists, they can serve as reliable second opinions, help reduce diagnostic errors and contribute to faster and more standardized evaluations.
Thus, the synergy between medicine and artificial intelligence offers exciting new possibilities for improving disease diagnosis, particularly in conditions like coeliac disease, and for delivering more effective, personalized care. The future of diagnostics may well be shaped by this alliance between human expertise and computational power, working together in the service of health.
Reference. Jaekle, F, Denholm, J, Schreiber, B, Evans, SC, Wicks, MN, Chan, JYH, Bateman, AC, Natu, S, Arends, MJ & Soilleux, E 2025, ‘Machine Learning Achieves Pathologist-Level Coeliac Disease Diagnosis’, NEJM AI. https://doi.org/10.1056/AIoa2400738
