Related Experiment Video
Updated: Jun 19, 2026
13:19
Deep Neural Networks for Image-Based Dietary Assessment
Published on: March 13, 2021
9.9K
Advanced deep feature engineering with crayfish optimization for diabetes detection using tongue images
Ghada Moh Samir Elhessewi1, Mukhtar Ghaleb2, Hany Mahgoub3
1Department of Health Sciences, College of Health and Rehabilitation Sciences, Princess Nourah Bint Abdulrahman University, P.O. Box 84428, 11671, Riyadh, Saudi Arabia.
Scientific Reports
|September 29, 2025
Summary
A new deep learning technique accurately detects diabetes mellitus (DM) using tongue images, offering a non-invasive alternative to traditional blood tests. This method achieves 96.91% accuracy, improving early disease prediction.
Area of Science:
- Biomedical imaging
- Medical diagnostics
- Artificial intelligence in healthcare
Background:
- Diabetes mellitus (DM) is a severe metabolic disease requiring early, non-invasive diagnostic methods.
- Current diagnostic approaches like fasting plasma glucose are invasive and time-consuming.
- Tongue image analysis presents a promising non-invasive avenue for DM detection.
Purpose of the Study:
- To develop an accurate, non-invasive method for diabetes mellitus detection using tongue imaging.
- To introduce the Deep Feature Engineering with Crayfish Optimization for Accurate Diabetes Disease Detection via Tongue Image Analysis (DFECO-DDTIA) technique.
- To enhance the accuracy and efficiency of diabetes diagnosis through advanced image analysis and optimization.
Main Methods:
- Image preprocessing using upgraded weighted median filtering (Up-WMF) for noise reduction.
- Feature extraction via squeeze-and-excitation-DenseNet (SE-DenseNet).
- Classification using temporal convolutional network (TCN) and hyperparameter tuning with Crayfish Optimisation Algorithm (COA).
Main Results:
- The DFECO-DDTIA technique achieved a high accuracy of 96.91% in diabetes detection from tongue images.
- The method demonstrated superior performance compared to existing diagnostic models.
- Image quality enhancement and optimized feature extraction contributed to improved diagnostic accuracy.
Conclusions:
- The DFECO-DDTIA technique offers a highly accurate and non-invasive approach for diabetes mellitus diagnosis.
- Deep learning and optimization algorithms show significant potential in biomedical imaging for disease detection.
- This study highlights the efficacy of tongue image analysis for early diabetes prediction.
