AI detects coronary heart disease from face and tongue photos in small test

A new artificial intelligence system developed by researchers in China has demonstrated the ability to accurately detect coronary heart disease (CHD) from photos of patients' faces and tongues, according to a recent study. The AI system, which utilizes deep learning to analyze images, identified every patient with CHD in a small test involving 518 participants - 200 diagnosed with CHD and 318 healthy volunteers.

The system performed separately on the face and tongue images, then combined the results for a more accurate prediction. The most successful version of the AI achieved 98% accuracy in testing 104 individuals, correctly identifying all 40 patients with CHD, though it made two false positive identifications of healthy participants as having the condition.

The AI system focused on specific areas of the tongue, surface coating, skin around the mouth and cheeks. The study highlights that certain physical features, such as darker tongues, thicker coatings, and facial pigmentation changes, are consistent with traditional Chinese medicine's beliefs about links between the tongue, face, and heart health.

While the AI showed promise in identifying CHD, it is important to note that the study had limitations. The participants were already diagnosed with CHD, so the AI was tested on a group with a higher risk of the condition. The images were taken under controlled conditions, and the study did not include a broader screening of the general population.

Additionally, false positives were identified, and further research is needed to establish the AI's ability to differentiate CHD from other conditions and to compare its effectiveness with existing risk assessment methods.