Can AI help detect diseases just by looking at your tongue pictures, eliminating the need for lab tests? Is it possible to receive an accurate diagnosis for your illnesses from the comfort of your home using just a smartphone? Will AI out-perform doctors in the near future?
A latest research titled: Tongue Disease Prediction Based on Machine Learning Algorithms, performed by researchers of university of Iraq and south Australia has left us wondering and questioning the future of in-hospital disease diagnosis by registered healthcare professionals. The research utilizes machine models for analyzing smartphone images of human tongue to detect several health conditions.
Understanding, how your tongue helps to detect illnesses
A healthy human's tongue typically appears pink with a thin white coating, owing to its textured surface due to presence of tiny white hair like papillae (structures present on tongue surface that are involved in taste perception).
Any changes in the colour and texture of tongue signal potential health concerns.
Red tongue may indicate vitamin deficiencies, infections, or particular conditions like radiation-induced oral mucositis or Kawasaki illness;
White tongue may indicate dehydration, oral thrush, or bacterial accumulation.
Black or dark tongue can develop as a result of pharmaceutical side effects, smoking, or underlying medical disorders including HIV.
A tongue that is yellow or purple indicates bacterial accumulation, liver difficulties, or circulation problems, respectively.
Grey or orange tongues can result from a variety of conditions, such as smoking, taking medication, or not brushing your teeth properly. Blue tongues can be a sign of low blood oxygen levels or certain medical problems, while green tongues may be caused by bacterial buildup or the intake of foods with pigment.
Being aware of these colour changes associated with tongue and what they indicate about your health can go a long way in helping to preserve your oral health and general well-being which further highlights the significance of routine medical and dental examinations in helping to identify and address these underlying health issues early on. [1]
Human tongues possess unique characteristics and features connected to the body’s internal organs, which effectively detect illnesses and monitor their progress. [2] We can therefore infer a lot about our health from our tongue. The tongue's colour and texture can reveal the underlying illnesses we are suffering from, long before we receive an actual diagnosis.
Based on this data, researchers from Australia and Iraq created an AI model that can assist in tongue analysis-based disease and condition detection. The AI model has a 98% accuracy rate in identifying the diseases just by analysing the pictures of the tongue, and detecting the changes in colour.
About the AI Model!!
The AI model was developed through a collaboration between Middle Technical University (MTU) in Baghdad and the University of South Australia (UniSA), reports Newswise.
Curious!! How this AI model functions?
According to Ali Al-Naji, the study is based on the conventional Chinese medical method of diagnosing illnesses, by looking at the tongue (tongue examination) and has been in use for over 2,000 years. In this practice, the doctor's analyses human tongue to check for diseases based on specific colours and textures indicating particular ailments. Leveraging this ancient technique with modern technology, the research team, led by Ali Al-Naji, an adjunct associate professor at both MTU and UniSA, sought to harness the power of AI to bring this diagnostic method into the 21st century.
Similar to other AI models, this one was trained by the researchers using a collection of 5,260 photographs of the tongue that had been labelled with outmost care, to match with different medical problems. Through training, the AI was able to learn, how to detect even minor changes in tongue texture and colour, the important markers of various health problems.
The research team further tested this AI model using 60 tongue photos from patients at two Middle Eastern teaching hospitals in order to confirm its correctness. The patients were seated around 20 cm (or 8 inches) from a laptop that had a webcam on it to take pictures of their tongues. After that, the AI model examined the photos and, almost always, correctly recognized the corresponding medical issues.
The AI model's precision in diagnosing illnesses
The XGBoost algorithm correctly predicted diseases 96.6% of the time during testing. This machine learning model is accurate because it continues improving itself after its initial output. It repeatedly makes a guess and calculates an error rate to gradually get closer to its goal, effectively training itself to increase its precision.
Unlike humans, this AI model isn’t limited to a narrow visible light spectrum and can accurately detect minor changes of saturation and luminance which boosts its diagnostic potential.
Lets look at the Advantages of this technology!
A single model can assist multiple patients simultaneously, additionally no clinic visit required, and it predicts diseases regardless of lighting, reducing the risk of inaccurate output.
At-home, AI-driven screening could revolutionize health care, making it more affordable and accessible. Millions of people pass away yearly from diseases they would’ve had a fighting chance against if they had caught it earlier.
AI-driven screening, at home could revolutionize healthcare by making it more timely, affordable and accessible, potentially reducing the number of deaths from diseases that could have been prevented if the diagnosis was received earlier.
AI's potential to revolutionize diagnostics could benefit hospitals and patients, by decreasing medical spending as it reduces the time spent by healthcare provider with patients.
AI's automation and autonomous nature could streamline medical appointments, which according to studies can help save 20-50% on the annual budgets, allowing hospitals to pass savings onto consumers or invest in lifesaving equipment.
Co-author of the study Javaan Chahl, noted that the AI model would soon be utilised as a smartphone app to diagnose a number of illnesses, including COVID-19, diabetes, stroke, anaemia, asthma, liver difficulties, and gallbladder disorders. Additionally, this will guarantee that the AI model is accessible to anyone.
These findings validate that computerised tongue analysis is a safe, effective, simple to use, and reasonably priced approach for disease screening that supports contemporary techniques with a centuries-old custom.
Javaan Chahl, Professor, UniSA
Reality Check!
The results of This research are not peer-reviewed which means they still have to be replicated by others to give it a thumbs up for wider approval and commercial utilization globally. Frankly it could take years before they use AI model analyzed tongue colors for diagnosis and disease prediction.
The researchers do admit that before this technology is widely used, there are still obstacles to be solved. Regulatory concerns, addressing patient concerns over data privacy and making sure that camera reflections don't affect the AI model's accuracy are two of the primary issues.
Can AI Models replace a doctor's diagnosis?
Most probably; AI may not replace doctors, but will certainly enhance disease identification, prediction, and treatment in the medical industry. Doctors globally, might begin utilizing the technology, on a larger scale, combining their expertise with machine learning power widely to boost their diagnostic potential.
Conclusion
The results of this study, which were published in the journal Technologies, showed that tongue analysis driven by AI has the potential to develop into a safe, effective, and approachable disease screening tool. In the future, according to the researchers, this technology might be included into a smartphone app that would let users snap a picture of themselves to get an instant health assessment.
Reference:
1. Hassoon, Ali Raad, Ali Al-Naji, Ghaidaa A. Khalid, and Javaan Chahl. 2024. “Tongue Disease Prediction Based on Machine Learning Algorithms.” Technologies 12 (7): 97. doi:10.3390/technologies12070097.
(Inputs from various sources)
(Rehash/Dr. Aditi Bakshi/MSM)
