Related Experiment Video
Updated: Oct 16, 2025
10:08
Co-culture of Glioblastoma Stem-like Cells on Patterned Neurons to Study Migration and Cellular Interactions
Published on: February 24, 2021
6.3K
Hierarchical-order multimodal interaction fusion network for grading gliomas
Man He1,2, Kangfu Han1,2, Yu Zhang1,2
1School of Biomedical Engineering, Southern Medical University, Guangzhou, Guangdong 510515, People's Republic of China.
Physics in Medicine and Biology
|October 19, 2021
Summary
This study introduces a novel deep learning method for grading brain gliomas using multimodal MRI. The approach achieves high accuracy in classifying tumor grades, aiding treatment planning.
Area of Science:
- Neuro-oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Gliomas are common primary brain tumors requiring accurate grading for treatment and prognosis.
- Multimodal magnetic resonance imaging (MRI) is crucial for glioma assessment.
Purpose of the Study:
- To develop a noninvasive deep learning method for glioma grading using multimodal MRI.
- To enhance multimodal fusion by leveraging collaborative and diverse high-order statistical information.
Main Methods:
- A novel high-order multimodal interaction module was designed for interactive learning and efficient fusion of multimodal data.
- High-order attention mechanisms were embedded to model complex statistical information and improve feature expression.
- Hierarchical recalibration of modality streams using diverse-order attention statistics was applied.
Main Results:
- The proposed deep learning method achieved high prediction performance on both BraTS2017 and TCIA datasets.
- Performance metrics included area under the ROC curve (95.2% on BraTS2017, 93.5% on TCIA), accuracy (94.28% on BraTS2017, 92.86% on TCIA), sensitivity, and specificity.
Conclusions:
- The developed deep learning approach demonstrates significant potential for accurate noninvasive glioma grading.
- Effective multimodal fusion using high-order statistical information and attention mechanisms enhances classification capabilities.
