A Sparse EEG-Informed fMRI Model for Hybrid EEG-fMRI Neurofeedback Prediction

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Science

Background:

• Neurofeedback (NF) utilizes brain activity measures like electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) for brain rehabilitation.
• While NF-EEG is established, NF-fMRI offers more precise brain training but is resource-intensive.
• Simultaneous EEG-fMRI neurofeedback (NF-EEG-fMRI) shows promise but faces patient burden challenges due to fMRI's nature.

Purpose of the Study:

• To develop a method for predicting bi-modal NF scores (NF-EEG-fMRI) using only EEG signals.
• To assess if EEG-only prediction can enhance brain rehabilitation protocols by reducing patient burden.
• To investigate the added value of predicting NF-fMRI scores from EEG compared to traditional NF-EEG.

Main Methods:

• A sparse regression model was developed to predict NF-fMRI and NF-EEG-fMRI scores from EEG recordings.
• The model was trained using simultaneous EEG and NF score data during motor imagery tasks.
• Performance was evaluated by comparing the correlation of predicted scores with actual bi-modal NF session outcomes.

Main Results:

• The proposed sparse regression model successfully predicted NF-fMRI and NF-EEG-fMRI scores using EEG data alone.
• Predicting NF-fMRI scores from EEG provided additional information beyond NF-EEG scores.
• The EEG-only prediction significantly improved correlation with bi-modal NF sessions compared to using NF-EEG scores alone.

Conclusions:

• Predicting bi-modal neurofeedback scores from EEG signals alone is feasible and effective.
• This EEG-only approach offers a promising, less burdensome alternative for brain rehabilitation protocols.
• Integrating predicted NF-fMRI information into EEG-based neurofeedback can enhance training efficacy.