AI model 98% accurate in detecting diseases from a tongue photo

This technology can be awesome!

Researchers in Iraq and Australia say they have developed a computer algorithm that can analyze the color of a person’s tongue to detect their medical condition in real time – with 98% accuracy.

“Typically, people with diabetes have a yellow tongue; cancer patients a purple tongue with a thick fatty layer; and acute stroke patients present with an unusually shaped red tongue,” explained senior study author Ali Al-Naji, who teaches at Baghdad Secondary Technical University and the University of South Australia.

“A white tongue can indicate anemia; people with severe cases of COVID-19 are likely to have a deep red tongue,” continued Al-Naji. “An indigo or purple tongue indicates vascular and gastrointestinal problems or asthma.”

Al-Naji says his proposed imaging system mimics the traditional Chinese medicine practice of examining the tongue for signs of disease.

More than 5,200 images were used to train the artificial intelligence model to identify the color of the tongue and the corresponding condition. The researchers tested it with 60 tongue images from two teaching hospitals in the Middle East.

Patients sat about 8 inches from a laptop equipped with a web camera, which photographed their tongue. The program was able to determine the disease in almost all cases.

The findings were published in the journal Technologies.

Study co-author Javaan Chahl, a professor at the University of South Australia, says the technology will eventually be used for a smartphone app that can diagnose diabetes, stroke, anemia, asthma, liver and gallbladder problems, COVID- 19 and other conditions.

“These results confirm that computerized language analysis is a safe, efficient, convenient and affordable method for disease screening that supports modern methods with a centuries-old practice,” Chahl said.

There are still some hurdles to overcome, including patient reluctance to provide data and camera reflections that fool the algorithm.

Also, a 2023 review of five years of analysis of AI language images raised concerns about researchers having to build their own datasets because there is no definitive dataset.

However, the technology was found to be of “enormous value” for diagnosing and treating disease.