Traditional Chinese medicine includes studying the tongue to detect illness. A healthy tongue appears pink with a thin white film. Not-so-healthy tongues may look white (lack of iron), bluish yellow (diabetes), purple (certain cancers), and so forth. Now, some researchers are working to adapt this diagnostic method by using AI technology.
Because color perception is very subjective and varying light conditions can affect color perception and analysis, researchers created a kiosk with a standardized lighting system. Patients would put their heads in the kiosk where images of their tongues could be collected.
Next, they trained an AI program to detect illnesses by showing it 5,260 images of healthy and unhealthy tongues, after which it
was able to classify and predict conditions such as diabetes, asthma, COVID,
and anemia, with 96.6 percent accuracy. After this training, the researchers
tested it with 60 tongue images they’d collected with their kiosk at two
hospitals. AI correctly identified patients’ conditions—based on their medical
records—in 58 out of 60 images.
According to the article in Scientific American, researchers
at the University of Missouri have launched an AI application, called BenCao, where
users can upload tongue images and receive personalized health guidance. The
app is designed and marketed as a “wellness” tool.
As you can imagine, the scientific community is wary about
studying the tongue to assess a patient’s health. On the other hand, tongue
examination is often part of a routine screening for oral cancer by dentists
and hygienists.
I’m guessing you’ll soon be sticking out your tongue in
front of a mirror.
For an introduction to this blog, see I Just Say No; for a list of blog topics, click the Topics tab.
