In a breakthrough that could soon reach UK hospitals, artificial intelligence (AI) systems are proving remarkably adept at diagnosing serious illnesses simply by analysing photographs of a patient's tongue. A major review of research has concluded these programs are accurate enough to assist doctors in spotting early signs of conditions like diabetes and gastric cancer.
How AI Decodes the Tongue's Secrets
The concept of tongue diagnosis is ancient, but the technology is cutting-edge. Scientists have developed AI programs that assess the tongue's colour, texture, and shape with impressive precision. These systems are 'trained' using vast databases containing thousands of tongue images from both sick and healthy individuals, learning to identify subtle visual patterns linked to specific diseases.
Professor Dong Xu, a bioinformatics expert at the University of Missouri, explains the process: "AI learns by identifying statistical patterns in large collections of tongue images paired with clinical data. It detects visual characteristics that appear more frequently in individuals with specific conditions." These characteristics include variations in colour distribution, surface texture, moisture, coating thickness, and swelling.
Striking Accuracy in Early Trials
The potential of this technology is underscored by compelling study results. In one 2024 study published in the journal Technologies, an AI program correctly identified 58 out of 60 patients with diabetes and anaemia based solely on a tongue picture.
Perhaps more significantly, research reported in eClinicalMedicine in 2023 found AI could spot gastric cancer from subtle changes in tongue colour and texture, such as a thicker coating or patchy redness linked to digestive tract inflammation. When tested, the AI distinguished cancer patients from healthy volunteers with 85 to 90 per cent accuracy – a rate comparable to standard diagnostic tests like gastroscopy or CT scans.
The Tongue as a Health Mirror
Medical experts affirm why the tongue is such a valuable diagnostic tool. "The tongue is referred to as the mirror of general health," says Professor Saman Warnakulasuriya, an emeritus professor of oral medicine at King's College London.
He details specific signs: a smooth, shiny tongue can indicate anaemia from nutrient deficiencies; a dry tongue may be an early diabetes symptom; while a thick white coating often signals infection. AI's advantage lies in its exhaustive training. While a GP may see limited examples, AI systems can recognise patterns clinicians might miss or that are invisible to the naked eye, prompting further investigation.
"The availability of clinical pictures in a well-trained AI program could give doctors confidence to narrow down a correct diagnosis," adds Professor Warnakulasuriya.
A Tool, Not a Replacement
Despite its promise, researchers caution that AI tongue analysis is an assistive tool, not a standalone diagnostician. The technology identifies visual correlations but doesn't understand underlying causes. For instance, a pale tongue flagged as a sign of anaemia could actually result from poor circulation.
Professor Bernhard Kainz of Imperial College London suggests the technology is most reliable as a broad health checker or triage tool. Professor Xu also warns that AI's effectiveness depends entirely on the quality and variety of its training data, and can be confused by factors like diet, hydration, or medication.
The consensus is clear: an AI tongue scan should never be treated as a final diagnosis. "It is always necessary to confirm the diagnosis by conducting appropriate laboratory tests," stresses Professor Warnakulasuriya. Used appropriately, however, this innovation could help prioritise patient care and reduce missed early warnings, complementing the irreplaceable clinical judgment of NHS doctors.
