Related Experiment Video
Updated: Dec 11, 2025
05:56
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
Published on: April 14, 2023
3.0K
TISNet-Enhanced Fully Convolutional Network with Encoder-Decoder Structure for Tongue Image Segmentation in
Xiaodong Huang1,2,3, Hui Zhang1,2, Li Zhuo1,2
1Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China.
Computational and Mathematical Methods in Medicine
|August 25, 2020
Summary
This study presents an enhanced deep learning method for accurate automated tongue image segmentation, crucial for digital tongue diagnostics. The novel approach significantly improves contour detection and pathological detail differentiation.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Automated tongue diagnosis relies on accurate tongue body segmentation from digital images.
- Challenges include blurred edges, pathological interference, and variations in tongue size and shape.
Purpose of the Study:
- To develop an automated tongue image segmentation method using an enhanced fully convolutional network (FCN) with an encoder-decoder structure.
- To improve the accuracy and reliability of tongue body extraction for automated tongue diagnoses.
Main Methods:
- Utilized a deep residual network as an encoder for dense feature maps.
- Incorporated a Receptive Field Block to capture global contextual information.
- Employed a Feature Pyramid Network as a decoder to fuse multi-scale features for precise contour recovery.
Main Results:
- Achieved high segmentation accuracy on the SIPL-tongue dataset with average Dice Similarity Coefficient of 97.26%.
- Demonstrated superior performance compared to SegNet, FCN, PSPNet, and DeepLab v3+.
- Validated effectiveness on the HIT-tongue dataset, meeting practical requirements for automated diagnosis.
Conclusions:
- The proposed enhanced FCN method provides accurate tongue image segmentation.
- This technique effectively addresses challenges in automated tongue diagnosis.
- The method shows significant potential for clinical application in digital health.
