A new study funded by the National Institute of Health and Care Research (NIHR) has shown that an artificial intelligence (AI) tool can accelerate the diagnosis of coeliac disease
Coeliac disease triggers the immune system to attack its own tissues when gluten is consumed, damaging the gut and preventing proper nutrient absorption. This leads to symptoms such as abdominal pain and bloating. Currently, diagnosis involves a blood test and, in some cases, a biopsy of the small intestine.
Now, research published in the New England Journal of Medicine has demonstrated that an AI tool can process biopsy samples more efficiently, reduce waiting times, and help specialist pathologists diagnose and treat more patients. The study was funded by the NIHR Invention for Innovation programme and led by Cambridge University Hospitals NHS Foundation Trust.
First use of AI in coeliac disease diagnosis
AI is widely used in healthcare, particularly in cancer diagnosis, but has never before been applied to coeliac disease.
Professor Liz Soilleux, study lead and Honorary Consultant Pathologist, stated: “It can take years to receive an accurate diagnosis, and with increasing pressure on healthcare systems, these delays are likely to continue. AI has the potential to speed up this process, allowing patients to receive a diagnosis sooner while also alleviating pressure on NHS waiting lists.”
The study utilised a type of AI known as machine learning. Researchers trained the tool on 4,000 biopsy images from five NHS hospitals to distinguish between healthy samples and those with coeliac disease. When tested on a further 650 biopsies, the AI achieved a diagnostic accuracy of over 97%, matching the performance of experienced pathologists.
Dr Florian Jaeckle, from the Department of Pathology and first author of the paper, explained: “This is the first time AI has been proven to diagnose coeliac disease as accurately as an experienced pathologist. Since we trained the model using data from various conditions, we are confident it can function effectively in different settings where biopsies are processed and imaged differently.”
“Our next step is to validate the algorithm on a much larger clinical sample, enabling us to present it to regulators and move closer to its implementation in the NHS.”
The University of Cambridge has established Lyzeum Ltd, a spinout company, to commercialise this AI tool.
Making AI diagnoses understandable
Following this breakthrough, researchers are collaborating with patient groups through Coeliac UK to improve the explainability of AI diagnoses. While AI detects patterns in data, understanding how these patterns translate into a disease diagnosis can be challenging. The team emphasises the importance of making AI-driven results transparent to build trust in its NHS adoption.
Liz Cox, 80, who has coeliac disease and serves as Secretary of a Coeliac UK support group, welcomed the development: “Anything that speeds up the system must be a good thing. Once you have a diagnosis and know you can’t eat gluten, you can take control of your health.”