Related Experiment Video
Updated: Jun 9, 2025
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
8.8K
Lightweight MRI Brain Tumor Segmentation Enhanced by Hierarchical Feature Fusion
Lei Zhang1, Rong Zhang1, Zhongjie Zhu1,2
1Ningbo Industrial Vision and Industrial Intelligence Lab, Zhejiang Wanli University, Ningbo 315100, China.
Tomography (Ann Arbor, Mich.)
|October 25, 2024
Summary
A new lightweight MRI brain tumor segmentation method, enhanced by hierarchical feature fusion (EHFF), reduces model parameters and improves boundary delineation. This approach achieves high accuracy on the BraTS 2021 dataset.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Current MRI brain tumor segmentation methods often have too many parameters.
- Existing techniques struggle with precise tumor boundary delineation.
Purpose of the Study:
- To introduce a lightweight MRI brain tumor segmentation method.
- To enhance segmentation performance and reduce model complexity.
Main Methods:
- Developed an enhanced hierarchical feature fusion (EHFF) method.
- Introduced adaptive feature learning (AFL) for macro perception and micro focus.
- Implemented hierarchical feature weighting (HFW) for multi-scale feature refinement.
- Utilized a hierarchical feature retention (HFR) module for detail preservation.
Main Results:
- The proposed EHFF method achieved superior performance on the BraTS 2021 dataset.
- Achieved Dice similarity coefficients (DSC) of 88.57% (ET), 91.53% (TC), and 93.09% (WT).
Conclusions:
- The EHFF method offers an effective and efficient solution for MRI brain tumor segmentation.
- This lightweight approach improves accuracy and detail preservation in tumor boundary segmentation.
Keywords:
MRI brain tumor segmentationhierarchical feature fusionlightweightmacro perceptionmicro focus
