Researchers from MTU and the University of South Australia (UniSA) explored how the color of a person’s tongue could signal a range of health conditions. The team trained machine learning models to interpret tongue hues as indicators of disease, aiming to offer a noninvasive diagnostic support tool. The study, described in Technologies magazine, underscores how AI can integrate traditional clinical observations with modern computation to assist medical assessment.
A collaboration between scientists in Iraq and Australia produced an imaging system designed to spot several health issues. The device analyzes tongue images to identify signs of diabetes, stroke risk, anemia, asthma, liver and gallbladder concerns, Covid-19, and a spectrum of vascular and gastrointestinal problems. The approach emphasizes color patterns as potential biomarkers that can be captured in standardized photographs and analyzed by algorithms.
To train the artificial intelligence, researchers compiled a dataset comprising 5260 tongue images from patients with diverse pathologies. After the preparation phase, the model was tested with 60 new photographs. The results showed the AI could classify the presence of disease with a remarkable 98 percent accuracy based on tongue color alone. The evaluation highlights the potential for rapid, noninvasive screening that could complement conventional diagnostics when deployed in appropriate clinical settings.
Findings noted that certain tongue colorations tended to align with specific conditions. For example, a yellow tint often appeared in individuals with diabetes, a purple tone correlated with cancer in some cases, and a red tongue was associated with a recent acute stroke. A pale or white tongue could signal anemia, while a dark red shade was observed in many Covid-19 cases under certain conditions. An indigo or purple hue was linked to asthma and also to issues with blood vessels, the stomach, or intestines. While these associations are not diagnostic on their own, they illustrate how tongue color could contribute to a broader clinical picture when used alongside other tests and observations.
The study’s lead author explained that the project drew inspiration from traditional Chinese medicine, which has long associated tongue appearance with health states. The researchers aimed to translate these ancient observations into a modern AI framework that can be consistently applied in contemporary medical workflows, moving from subjective interpretation to data-driven assessment.
Historically, medical practitioners have used tongue examination as a supplementary diagnostic tool. The new AI-based approach seeks to systematize and quantify those cues, reducing variability between observers and enabling scalable analysis. The technology is envisioned as an adjunct to standard diagnostic processes, helping clinicians triage patients, prioritize testing, and monitor disease progression through periodic tongue imagery.