Related Experiment Video
Updated: Sep 16, 2025
05:56
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
Published on: April 14, 2023
2.6K
GA-TongueNet: tongue image segmentation network using innovative DiFP and MDi for stable generalization ability
Zhiyu Dong1, Le Zhao1, Yajun Fan1
1College of Information Science and Engineering, Henan University of Technology, Zhengzhou, China.
Frontiers in Physiology
|July 9, 2025
Summary
A new AI model, GA-TongueNet, improves tongue image segmentation for Traditional Chinese Medicine diagnosis. It achieves high accuracy and stable generalization even with limited data and complex backgrounds.
Area of Science:
- Computer-aided diagnosis
- Medical imaging analysis
- Traditional Chinese Medicine (TCM)
Background:
- Tongue diagnosis in TCM is crucial for internal organ assessment.
- Accurate tongue image segmentation is vital for computer-aided diagnosis.
- Current algorithms struggle with small sample sizes and complex backgrounds, limiting practicality.
Purpose of the Study:
- To develop a robust tongue segmentation network with strong generalization ability and accuracy.
- To address limitations of existing algorithms in small sample and diverse background scenarios.
Main Methods:
- Proposed GA-TongueNet, a transformer-based network incorporating dilated feature pyramid (DiFP) and multi-dilated convolution (MDi) modules.
- DiFP module captures both global structure and local details.
- MDi module preserves high feature resolution, enabling capture of long-range dependencies and semantic content while retaining low-level details.
Main Results:
- GA-TongueNet demonstrates superior accuracy and generalization compared to existing CNN and Transformer-based semantic segmentation algorithms.
- The network maintains precision and stability even with limited sample sizes and complex imaging conditions.
- Experimental results validate the effectiveness of the proposed DiFP and MDi modules.
Conclusions:
- GA-TongueNet offers a significant advancement in tongue image segmentation for computer-aided TCM diagnosis.
- The proposed architecture effectively handles challenges of small sample sizes and complex backgrounds.
- This approach enhances the practicality and reliability of automated tongue diagnosis systems.
