Related Experiment Video
Updated: Sep 11, 2025
10:25
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
48.4K
Leveraging Prior Knowledge in a Hybrid Network for Multimodal Brain Tumor Segmentation
Gangyi Zhou1, Xiaowei Li1, Hongran Zeng1
1College of Electronics and Information Engineering, Sichuan University, Chengdu 610017, China.
Sensors (Basel, Switzerland)
|August 14, 2025
Summary
A new Hybrid Network for Multimodal Brain Tumor Segmentation (HN-MBTS) improves deep learning accuracy for MRI analysis. This advanced method enhances clinical diagnosis and treatment planning for brain tumors.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neuroscience
Background:
- Deep learning advances brain tumor segmentation in MRI, aiding clinical decisions.
- Challenges remain in integrating medical knowledge, multimodal feature capture, and boundary delineation.
Purpose of the Study:
- To introduce the Hybrid Network for Multimodal Brain Tumor Segmentation (HN-MBTS).
- To enhance brain tumor segmentation by integrating prior medical knowledge and improving feature extraction and boundary precision.
Main Methods:
- Proposed HN-MBTS incorporates prior medical knowledge for refined feature extraction and boundary precision.
- Key modules include Two-Branch, Two-Model Attention (TB-TMA) for multimodal fusion, Linear Attention Mamba (LAM) for multi-scale modeling, and Residual Attention (RA) for boundary refinement.
Main Results:
- The HN-MBTS method significantly outperforms existing approaches in brain tumor segmentation.
- Achieved average Dice scores of 87.66% on BraT2020 and 88.07% on BraT2023 datasets.
- Demonstrated superior segmentation accuracy and efficiency.
Conclusions:
- The proposed HN-MBTS offers enhanced accuracy and efficiency for multimodal brain tumor segmentation.
- This method shows significant potential for clinical applications in diagnosis and treatment planning.
