Using artificial intelligence to improve diagnosis of coeliac disease by biopsy

Project Title: An artificial intelligence solution for diagnosis and assessment of small intestinal biopsies in suspected or known coeliac disease

Professor Elizabeth Soilleux, is an academic clinical pathologist at the University of Cambridge and honorary consultant at Addenbrooke’s hospital and this project brought together a multidisciplinary team of pathologists and mathematicians from the University of Cambridge and the University of Edinburgh.

This study aimed to use biopsy images of confirmed coeliac disease and biopsy images of ‘normal’ small intestines to develop an artificial intelligence algorithm that could accurately distinguish biopsy images from a person with coeliac disease from biopsy images from people without coeliac disease.

Currently, if a person is suspected to have coeliac disease, a doctor will request a blood test to look for the antibodies associated with the condition.If the blood test is positive, the diagnosis may be confirmed either by a second blood test orby an endoscopy and biopsy. This is where a small tube with a camera on the end is passed through the mouth into the small intestine and pieces of tissue lining the gut are removed for further analysis. A trained pathologist looks at these tissue samples under a microscope to look for signs of damage associated with coeliac disease. Unfortunately, due to a shortage of specialist pathologists in the UKNHS, this process can sometimes take too long. Not only this, but it’s though that there could be a difference of opinions of up to 25%between trained expert pathologists.

In 2019, along with Innovate UK, we awarded £175k of funding to a project aiming to develop an artificial intelligence software system that could look at these small intestine biopsies and identify if the person it came from has coeliac disease.

Artificial intelligence can be described as a computer system that uses information to learn and to recognise patterns independently, without human interference. Artificial intelligence means the reproduction of human intelligence in computer automated machines. Through complex programming, the machine learns how to process information and make decisions based on the information received, similar to the way humans do.

The project was a success, and the funding facilitated the development of a software system that is able to identify coeliac disease biopsies with an accuracy of 95%, which is a greater accuracy than that of trained pathologists (85%). The National Institute for Health and Care Research (NIHR) has awarded the team a further £620k worth of funding to continue this groundbreaking research. That means that the return on investment is about £3.60 for every £1originally awarded by Coeliac UK and Innovate UK. The research team intend to use the follow on funding to improve the intelligence of the system so that it’s able to identify coeliac disease biopsies amongst a variety of other gastrointestinal conditions. They also plan to refine the software to hopefully improve accuracy further. They will then seek to validate the system in the hope that one day, it might be ready for clinical use, to allow the majority of people to be diagnosed with coeliac disease much quicker than if they had to wait for a pathologist to be available. Uncertain cases would still need to be looked at by an expert pathologist.

The original project was jointly funded by a grant from Coeliac UK and Innovate UK. Innovate UK is an organisation within the UK government’s national research and funding agency that invests in science and research.