Related Experiment Video
Updated: May 8, 2026
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
2.2K
WSSM: A Weakly Supervised Oral Mucosal Disease Segmentation Model Based on Multi-Task Collaboration
Jing Xu1, Jianguo Ju2, Qian Zhang1
1College of Computer Science, Northwest University, Xi'an, China.
Oral Diseases
|January 13, 2026
Summary
This study introduces a weakly supervised oral mucosal disease (OMD) segmentation model (WSSM) that enhances diagnostic accuracy. WSSM improves lesion boundary segmentation, addressing limitations in traditional OMD diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Traditional oral mucosal disease (OMD) diagnosis is subjective and inefficient, relying on visual assessment.
- OMD imaging lacks sufficient supervision and has fuzzy lesion boundaries, hindering mobile medicine applications.
Purpose of the Study:
- To develop a weakly supervised OMD segmentation model (WSSM) to improve diagnostic accuracy and efficiency.
- To address challenges of insufficient supervision and fuzzy boundaries in OMD image analysis.
Main Methods:
- Proposed a weakly supervised OMD segmentation model (WSSM) with a Mamba backbone and dual-branch collaboration.
- Implemented a classification branch for multi-scale feature extraction and a pseudo-label module for deeper supervision.
- Utilized a segmentation branch with a boundary adaptive module to enhance fuzzy boundary representation.
Main Results:
- WSSM significantly outperformed existing weakly supervised methods on the OMD dataset.
- Achieved a 6.06% increase in Dice index compared to WSSL, demonstrating superior segmentation performance.
Conclusions:
- The Mamba-based WSSM balances local texture and long-range dependencies for OMD lesion analysis.
- Dual-branch collaboration and deeper supervision significantly improve boundary segmentation accuracy for OMDs with unclear boundaries.
