Related Experiment Video
Updated: Aug 5, 2026
11:28
Concurrent EEG and Functional MRI Recording and Integration Analysis for Dynamic Cortical Activity Imaging
Published on: June 30, 2018
Interpretable Dual-Stream EEG-MRI Fusion Uncovers Structure-Function Signatures of Stroke Motor Recovery
Summary
Predicting motor recovery after stroke is improved by a new deep learning framework that integrates electroencephalography (EEG) and structural magnetic resonance imaging (MRI). This interpretable model captures structure-function interactions for better stroke outcome prediction.
Area of Science:
- Neuroscience
- Medical Imaging
- Artificial Intelligence
Background:
- Predicting motor recovery post-stroke is challenging due to limitations in single-modality imaging.
- Existing multimodal approaches often fail to capture the interplay between structural damage and functional reorganization.
- Current methods treat electroencephalography (EEG) and magnetic resonance imaging (MRI) signals independently, limiting predictive power and interpretability.
Purpose of the Study:
- To introduce Dual-Stream Cross-Modal Fusion (DS-CMF), an interpretable deep learning framework.
- To model bidirectional structure-function coupling by integrating EEG and structural MRI data.
- To improve motor recovery prediction and mechanistic understanding in subacute ischemic stroke patients.
Main Methods:
- Developed DS-CMF, a deep learning framework for patient-adaptive integration of EEG and structural MRI.
- Fused task-based EEG connectivity features with structural MRI descriptors using region-aligned embedding and bidirectional cross-modal attention.
- Applied the framework to 26 subacute ischemic stroke patients performing motor imagery tasks (KGMI, PGMI, WEMI, WFMI) and used SHapley Additive exPlanations (SHAP) for feature attribution.
Main Results:
- DS-CMF achieved high accuracies in predicting motor tasks, notably 77.3% for power grip imagery (PGMI) and 73.3% for wrist extension imagery (WEMI).
- The framework demonstrated incremental value over simpler fusion methods, particularly for PGMI and WEMI.
- SHAP attribution identified potential structure-function markers, including specific EEG connectivity patterns and structural contributions from sensorimotor and frontal regions.
Conclusions:
- Interpretable fusion of EEG and structural MRI shows promise for capturing relevant structure-function patterns in subacute stroke.
- DS-CMF offers a proof-of-concept for enhanced motor recovery prediction and mechanistic insights.
- Further validation is needed to confirm identified features as reliable biomarkers for stroke recovery.
