Related Experiment Video
Updated: Jul 8, 2026
12:03
Multi-photon Imaging of Tumor Cell Invasion in an Orthotopic Mouse Model of Oral Squamous Cell Carcinoma
Published on: July 25, 2011
19.3K
NTSM: a non-salient target segmentation model for oral mucosal diseases
Jianguo Ju1, Qian Zhang1, Ziyu Guan1
1School of Information Science and Technology, Northwest University, No.1, Xuefu Road, Xi'an, 710119, Shaanxi, China.
BMC Oral Health
|May 2, 2024
Summary
A new non-salient target segmentation model (NTSM) improves oral lesion segmentation accuracy. This model significantly reduces parameters, enhancing deployability for medical imaging devices.
Area of Science:
- Medical image analysis
- Computer vision in healthcare
- Biomedical engineering
Background:
- Accurate segmentation of oral mucosal diseases is challenging due to non-salient features similar to normal tissues.
- High-precision models often have excessive parameters, hindering deployment on portable devices.
Purpose of the Study:
- To develop a non-salient target segmentation model (NTSM) that enhances segmentation performance while reducing parameter count.
- To address the limitations of existing models in segmenting subtle oral lesions and resource-intensive deployments.
Main Methods:
- The NTSM incorporates a difference association (DA) module to enhance feature differences and local context.
- Multiple feature hierarchy pyramid attention (FHPA) modules are utilized to extract pathological information efficiently.
- The DA module learns semantic relationships using varied receptive fields to improve lesion identification.
Main Results:
- The NTSM demonstrated superior performance on oral mucosal diseases (OMD) and international skin imaging collaboration (ISIC) datasets.
- Compared to nnU-Net, the proposed model reduced parameters by 43.20% while increasing the Dice score by 3.14%.
Conclusions:
- The developed model achieves high segmentation accuracy for non-salient oral mucosal disease areas.
- The NTSM effectively reduces computational resource consumption, facilitating practical application.
