98% accurate disease prediction by AI based on tongue color analysis

Over the past three years, healthcare professionals have harnessed artificial intelligence (AI) and used it to quickly and accurately diagnose diseases in patients.

From over 90 percent accuracy in AI prediction of lung disease to general AI models that predict disease potential even before conventional diagnoses have been made, technology has advanced rapidly today.

Now, engineering researchers at the Middle Technical University (MTU) in Baghdad, Iraq, and the University of South Australia (UniSA) in Adelaide, Australia, have made a breakthrough in training machine learning algorithms to detect diseases by analyzing tongue color.

An accuracy of 98 percent was achieved in predicting various diseases when the AI ​​algorithm examined the color of the human tongue.

AI diagnoses diabetes, stroke, COVID-19

According to a statement from the scientists, the AI ​​model diagnosed diabetes, stroke, anemia, asthma, liver and gallbladder diseases, COVID-19, and a number of vascular and gastrointestinal problems.

Ali Al-Naji, lead author at MTU and associate professor at UniSA, explained that the AI ​​mimics a 2,000-year-old practice widely used in traditional Chinese medicine – examining the tongue for signs of disease.

“The color, shape and thickness of the tongue can reveal a whole range of health conditions,” he added.

“Typically, people with diabetes have a yellow tongue, cancer patients have a purple tongue with a thick, fatty coating, and patients with an acute stroke have an unusually shaped red tongue.”

Al-Naji also pointed out that a white tongue can indicate anemia. People with severe cases of COVID-19 are likely to have a deep red tongue. An indigo or purple-colored tongue indicates vascular and gastrointestinal problems or asthma.

AI model trained with 5260 images

Therefore, experts trained the computer vision systems equipped with a new imaging system using 5260 images characterized by the visible color classes red, yellow, green, blue, gray, white and pink.

Using six machine learning algorithms, the AI-powered computer algorithms were trained to predict tongue color in all lighting conditions.

These systems are – the methods Naïve Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Decision Trees (DTs), Random Forest (RF) and Extreme Gradient Boost (XGBoost),

The authors stated in the study that the research proposes a new AI imaging system to analyze and extract tongue color features at different color saturations and under different lighting conditions from five color space models (RGB, YcbCr, HSV, LAB, and YIQ).

To test the system, 60 images of the tongues of patients with specific health conditions were obtained from two teaching hospitals in the Middle East.

In most cases, the AI ​​model successfully matched the tongue color to the diseases diagnosed in these patients, the statement said.

In order to capture images of the patient’s tongue, the cameras were positioned an additional 20 centimeters away from the fleshy muscle organ.

Real-time diagnostics

This confirmed that AI could indeed advance the medical field by detecting diseases faster and enabling instant diagnosis by closely examining the color of the tongue.

Real-time diagnosis could reduce waiting times in hospitals and speed up queues. While a human might be required to confirm patients’ illnesses, the AI ​​model would help professionals confirm the diagnosis faster.

Javaan Chahl, co-author from UniSA and professor, explained that smartphones could one day be used in this way to diagnose diseases.

“These results confirm that computer-assisted tongue analysis is a safe, efficient, user-friendly and cost-effective method for disease detection that complements modern methods with a centuries-old practice,” he says.

The study was published in the journal – Technologies.