- Traditional tongue disease diagnosis is subjective and inconsistent, especially in TCM.
- AI-driven systems offer improved accuracy and reliability in tongue diagnostics.
- Researchers developed a machine-learning-based imaging system to analyze tongue color under various conditions.
- The system uses six machine learning algorithms, with XGBoost performing at 98.7% accuracy.
- The study included 5,260 training images and 60 pathological images for testing.
- The XGBoost-based system showed 96.6% accuracy in real-time disease prediction.
- Different tongue colors correspond to specific health conditions, e.g., yellow for diabetes and blue for asthma.
- The system is user-friendly, efficient, and cost-effective, but future improvements are needed for image processing.
Main AI News:
In a groundbreaking study, researchers introduced a novel machine-learning-based imaging system to analyze and extract tongue color features under different light conditions and color saturations for real-time disease prediction. This innovative system utilized six machine learning algorithms, including support vector machines (SVM), naive Bayes (NB), decision trees (DTs), k-nearest neighbors (KNN), Extreme Gradient Boost (XGBoost), and random forest (RF) classifiers to predict tongue color.
The study employed various color models, such as the Human Visual System (HSV), the red, green, and blue system (RGB), luminance separation from chrominance (YCbCr, YIQ), and lightness with green-red and blue-yellow axes (LAB), to enhance the precision of color analysis.
The researchers divided their data into training (80%) and testing (20%) sets. The training set included 5,260 tongue images, categorized by color—yellow, red, blue, green, pink, white, and gray—under various light conditions and saturations. Additionally, the study involved 60 pathological tongue images from Mosul General Hospital and Al-Hussein Hospital in Iraq, featuring patients with conditions like diabetes, asthma, mycotic infections, kidney failure, COVID-19, anemia, and fungiform papillae.
During the testing phase, patients sat 20cm from the camera while the machine-learning algorithms analyzed tongue colors and predicted health status in real-time. The imaging system used MATLAB App Designer on laptops and high-resolution webcams to extract and analyze tongue images, isolating the central tongue region for accurate assessment.
The system then converted RGB images into various color models (HSV, YCbCr, YIQ, and LAB) and processed the data through multiple machine learning algorithms for model training. Performance was evaluated using precision, accuracy, recall, F1-scores, and others.
XGBoost emerged as the top performer, achieving an impressive accuracy rate of 98.7%, while the naive Bayes algorithm lagged with 91% accuracy. XGBoost also excelled in other metrics, such as F1 scores (98%), Jaccard index (0.99), G-score (0.92), and others, demonstrating its high reliability for tongue analysis. Consequently, the researchers chose XGBoost as the final imaging tool, which predicts tongue color and related diseases via a real-time, user-friendly graphical interface.
The system delivered strong results upon deployment, correctly identifying 58 out of 60 tongue images with a 96.6% accuracy rate. Notably, a pink tongue indicated good health, while other colors signaled various illnesses: yellow for diabetes, green for mycotic diseases, blue for asthma, red for COVID-19, black for fungiform papillae, and white for anemia.
Conclusion
Introducing AI-driven systems like the one discussed marks a significant development in the healthcare diagnostics market. With its high accuracy and real-time capabilities, this technology is set to disrupt traditional diagnostic methods, particularly in areas like TCM, where subjectivity has been a challenge. The ability to deploy this system in real-world scenarios with impressive results indicates a strong potential for widespread adoption, driving demand for AI-integrated diagnostic tools. This shift towards AI in diagnostics could also spur innovation in other areas of medical imaging, offering new growth opportunities for tech companies in the healthcare sector. However, addressing the current limitations, such as camera reflections and the need for enhanced image processing, will be crucial for maintaining competitive advantage and ensuring sustained market growth.
