โก Quick Summary
The study introduces a novel weakly supervised oral mucosal disease (OMD) segmentation model called WSSM, which utilizes multi-task collaboration to enhance lesion segmentation accuracy. This model significantly outperformed existing methods, achieving a 6.06% increase in the Dice index compared to WSSL.
๐ Key Details
- ๐ Dataset: OMD dataset
- ๐งฉ Features used: Multi-directional, multi-scale features
- โ๏ธ Technology: WSSM with Mamba as the backbone
- ๐ Performance: Dice index increased by 6.06% compared to WSSL
๐ Key Takeaways
- ๐ง Traditional OMD diagnosis is often subjective and inefficient.
- ๐ก WSSM employs a dual-branch collaboration for improved segmentation.
- ๐ The classification branch captures essential features through a dedicated network.
- ๐ ๏ธ Pseudo-label module enhances supervision by fusing various annotations.
- ๐ Boundary adaptive module improves representation of fuzzy lesion boundaries.
- ๐ Results indicate significant advancements in segmentation accuracy.
- ๐ Potential applications in mobile medicine and telehealth.
- ๐ Paper Code: Available at GitHub.
๐ Background
The diagnosis of oral mucosal diseases has traditionally relied on the expertise of clinicians, which can lead to high subjectivity and inefficiency. The challenge is compounded by the lack of sufficient supervision information and the presence of fuzzy lesion boundaries in OMD images. As mobile medicine continues to evolve, the demand for high accuracy in diagnosis becomes increasingly critical.
๐๏ธ Study
The researchers developed the WSSM model to address the limitations of traditional OMD diagnosis. By leveraging the Mamba backbone, the model integrates a classification-segmentation dual-branch approach. The classification branch is designed to capture multi-directional and multi-scale features, while the segmentation branch focuses on extracting overall lesion features, enhanced by a boundary adaptive module for improved accuracy.
๐ Results
The experiments conducted on the OMD dataset revealed that WSSM significantly outperformed existing weakly supervised methods. The model achieved a Dice index increase of 6.06% compared to WSSL, demonstrating its effectiveness in enhancing segmentation accuracy, particularly in scenarios with insufficient supervision and unclear boundaries.
๐ Impact and Implications
The introduction of the WSSM model could have profound implications for the field of oral medicine. By improving the accuracy of OMD segmentation, this model paves the way for more reliable diagnoses and treatment plans. The advancements in segmentation technology could also enhance the capabilities of mobile medicine, making it easier for healthcare providers to deliver accurate assessments remotely.
๐ฎ Conclusion
The WSSM model represents a significant breakthrough in the segmentation of oral mucosal diseases. By utilizing a multi-task collaborative approach, it addresses the challenges of traditional diagnostic methods, offering a promising solution for improving accuracy in OMD diagnosis. Continued research and development in this area could lead to even greater advancements in the integration of AI technologies in healthcare.
๐ฌ Your comments
What are your thoughts on the advancements in OMD segmentation technology? We would love to hear your insights! ๐ฌ Leave your comments below or connect with us on social media:
WSSM: A Weakly Supervised Oral Mucosal Disease Segmentation Model Based on Multi-Task Collaboration.
Abstract
BACKGROUND: Traditional oral mucosal disease (OMD) diagnosis relies heavily on clinicians’ experience and visual assessment, suffering from high subjectivity and low efficiency. OMD images also have insufficient supervision information and fuzzy lesion boundaries, failing to meet mobile medicine’s high accuracy requirements.
METHODS: To solve these issues, we proposed a weakly supervised OMD segmentation model with multi-task collaboration (WSSM). Using Mamba as the backbone, WSSM realizes efficient lesion segmentation via classification-segmentation dual-branch collaboration. The classification branch captures multi-directional, multi-scale features via a dedicated network, and its pseudo-label module fuses class activation maps, box annotations, and predictive annotations for deeper supervision. The segmentation branch adopts a symmetric network to extract overall lesion features, with a boundary adaptive module enhancing fuzzy boundary representation to improve accuracy.
RESULTS: Experiments on the OMD dataset demonstrated that WSSM outperformed existing weakly supervised methods significantly, with its Dice index increasing by 6.06% compared to WSSL.
CONCLUSIONS: Our model, with Mamba as the backbone (balancing local texture feature extraction and long-range semantic dependency modeling of OMD lesions), enables deeper supervision via dual-branch collaboration, significantly improving boundary segmentation accuracy in scenarios with insufficient OMD supervision and unclear boundaries. PAPER CODE: https://github.com/XJ156/WSSM3.git.
Author: [‘Xu J’, ‘Ju J’, ‘Zhang Q’, ‘Shen X’, ‘Zeng X’, ‘Nurmanbai CL’, ‘Guan Z’, ‘Shen Z’, ‘Xu P’]
Journal: Oral Dis
Citation: Xu J, et al. WSSM: A Weakly Supervised Oral Mucosal Disease Segmentation Model Based on Multi-Task Collaboration. WSSM: A Weakly Supervised Oral Mucosal Disease Segmentation Model Based on Multi-Task Collaboration. 2026; (unknown volume):(unknown pages). doi: 10.1111/odi.70204