AI Can Spot Diabetes & Stomach Cancer from Tongue Color, Studies Show

These programs are trained on databases containing thousands of tongue images from patients with known illnesses. By learning visual patterns linked to disease, the systems assess subtle changes that may not be obvious to clinicians.

Doctors have long examined the tongue for clues to underlying illness, using color, texture, and moisture as visible indicators of health. Now, scientists say artificial intelligence may be able to do the same, faster, and with striking accuracy.

A review of more than 20 studies has found that AI programs analyzing tongue images can identify early signs of conditions, including diabetes and stomach cancer.

The findings, reported in the journal Chinese Medicine, suggest the technology could soon be used in hospitals to support diagnosis.

In one of the most notable studies, published in 2024 in the journal Technologies, an AI system correctly diagnosed 58 out of 60 patients with diabetes and anaemia using only photographs of their tongues.

These programs are trained on databases containing thousands of tongue images from patients with known illnesses. By learning visual patterns linked to disease, the systems assess subtle changes that may not be obvious to clinicians.

Another study showed that AI could detect gastric cancer from tongue features such as thicker coatings, patchy colour loss and areas of redness associated with digestive inflammation.

When tested on new patients, the system distinguished cancer cases from healthy volunteers with accuracy comparable to gastroscopy or CT scans, identifying cases correctly around 85 to 90 per cent of the time, according to eClinicalMedicine in 2023.

‘AI learns by identifying statistical patterns in large collections of tongue images paired with [the patient’s] clinical or health-related data,’ said Professor Dong Xu, a bioinformatics expert at the University of Missouri. ‘It detects visual characteristics that appear more frequently in individuals with specific conditions than in healthy people.’

Experts say the concept itself is not new. ‘The tongue is referred to as the mirror of general health,’ said Saman Warnakulasuriya, emeritus professor of oral medicine and experimental pathology at King’s College London.

He explained that a smooth tongue can indicate anaemia due to deficiencies in iron, vitamin B12 or folate, while dryness may signal diabetes-related dehydration and nerve damage affecting saliva production. High blood sugar can also encourage bacterial or fungal growth, leading to yellowish coatings.

While AI systems may identify patterns missed in routine practice, experts stress that the technology has limits. A pale tongue, for example, could reflect factors other than anaemia, such as poor circulation.

AI also lacks the ability to assess full medical histories, symptoms or lifestyle factors. ‘The availability of clinical pictures in a well-trained AI program could give doctors confidence to narrow down a correct diagnosis,’ Professor Warnakulasuriya said.

Variations in lighting, camera quality, hydration, diet, smoking and medication can all influence tongue appearance, Professor Xu added, warning that AI performance depends heavily on the data it is trained on.

Used appropriately, researchers say, AI tongue analysis could help prioritise care and reduce missed early signs of disease, while remaining a support tool, not a replacement for clinical judgment.

Stay tuned for more such updates on Digital Health News