AI Tool Detects Coronary Heart Disease from Face and Tongue Images

Coronary heart disease (CHD) develops when fatty plaques narrow the arteries supplying the heart muscle, restricting blood flow and sometimes first presenting as chest pain, myocardial infarction, or another serious cardiovascular event. Earlier identification therefore remains an important goal in preventive care. Supporting this effort, researchers in China have developed an artificial intelligence approach that detects CHD from photographs of the face and tongue.

The technology is based on the Smart TCM-A1 system, which captures standardized face and tongue images for analysis. In the study from China, the system used deep learning to assess visible image patterns associated with diagnosed CHD. It analyzed tongue and face photographs separately before combining both sets of information to predict disease status.

The researchers recruited 518 participants, including 200 people with diagnosed CHD and 318 healthy volunteers. Each participant was photographed using a purpose-built scanning system under controlled conditions. The combined face-and-tongue model performed better than either image type alone.

In a test group of 104 people, the best-performing version was 98% accurate. It correctly identified all 40 patients with CHD. It also incorrectly classified two healthy participants as having the condition, producing two false-positive results.

The AI model focused on the center of the tongue, tongue coating, skin around the mouth, and cheeks. Some patients with CHD showed darker tongues, thicker coatings, or changes in facial pigmentation. The researchers reported that these highlighted regions were consistent with traditional Chinese medicine concepts linking the tongue, face, and heart.

The study was published in Scientific Reports on September 18, 2026. Its findings suggest that a rapid photographic test could potentially support decisions about which patients need further cardiovascular evaluation. Current CHD assessment relies on risk factors such as age, smoking, blood pressure, cholesterol, and diabetes, with suspected cases evaluated using tools such as electrocardiography, blood tests, and specialized imaging.

Important questions remain before the approach can be used clinically. The study compared patients with established CHD against healthy volunteers, which differs from screening a general population with varied cardiovascular risk. The model has not yet been tested across other hospitals, ethnic groups, or care settings, and larger longitudinal studies would be needed to determine whether it can detect previously unrecognized disease and improve patient outcomes.