Related Experiment Video
Updated: Jul 5, 2026
09:41
A Pipeline for 3D Multimodality Image Integration and Computer-assisted Planning in Epilepsy Surgery
Published on: May 20, 2016
PF-CMNet: Progressive Frequency-Aware Cross-Modal Network with Missing-Modality Distillation for 3D Brain Tumor
Haokun Wang1, Shuyi Wang1, Yuqi Li1
1School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China.
Brain Sciences
|June 26, 2026
Summary
This study introduces PF-CMNet, a novel framework for brain tumor segmentation using multimodal MRI. The model demonstrates improved accuracy and robustness, even with incomplete imaging data.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neuroscience
Background:
- Accurate segmentation of multimodal MRI is crucial for neurosurgery.
- Existing models struggle with low contrast, ambiguous boundaries, and missing imaging data.
Purpose of the Study:
- Develop a robust segmentation framework for multimodal MRI.
- Improve cross-modal learning, boundary recovery, and performance with incomplete data.
Main Methods:
- Proposed PF-CMNet: Progressive Frequency-Aware Cross-Modal Network with Missing-Modality Distillation.
- Utilized Cross-Modal Selective Frequency Attention and Progressive Cross-Scale Detail Fusion.
- Employed teacher-student distillation for robustness against missing modalities.
Main Results:
- Achieved 84.3% average Dice score on MSD Task01_BrainTumour.
- Attained 88.2% average Dice score on BraTS2021, with the lowest Hausdorff distance.
- Maintained strong performance with missing MRI sequences in stress tests.
Conclusions:
- PF-CMNet offers a unified framework for multimodal brain tumor segmentation.
- The model enhances accuracy, boundary consistency, and robustness to incomplete MRI.
- Achieved a favorable accuracy-efficiency trade-off.
