Related Experiment Video
Updated: Jul 14, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
TUMamba: A novel tongue segment methods based on Mamba and U-Net
Fan Jiang1, Yanmei Zhong2, Simin Yang3
1Rehabilitation Medicine Department, The first Affiliated Hospital of Chongqing Medical University, Chongqing, China.
Digital Health
|November 1, 2024
Summary
This study introduces a novel tongue segmentation model combining Mamba and U-Net to improve accuracy in complex environments. The TUMamba model enhances feature extraction and filtering, outperforming existing methods.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Current tongue segmentation methods face challenges with global feature extraction and selective filtering in complex environments.
- Background objects resembling the tongue can significantly reduce segmentation efficiency and accuracy.
Purpose of the Study:
- To propose a novel model for tongue segmentation in complex environments.
- To enhance segmentation accuracy and efficiency by combining Mamba and U-Net.
Main Methods:
- Incorporated the Mamba attention module and a multi-stage feature fusion module into the U-Net backbone.
- Mamba attention module serially connects spatial and channel attention at U-Net's skip connections for selective feature map filtering.
- Multi-stage feature fusion module integrates feature maps from different stages to improve performance.
Main Results:
- The proposed model achieved a 1.17% improvement in mean intersection over union (mIoU) compared to state-of-the-art models.
- Ablation experiments confirmed that each proposed module contributes to enhanced segmentation efficiency.
Conclusions:
- The Tongue segmentation model based on U-Net and Mamba (TUMamba) effectively extracts global and local features.
- The model excels in tongue segmentation tasks within complex environments, demonstrating its practical value.
