A new informed tensor factorization approach to EEG-fMRI fusion

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• The relationship between synchronous neuronal activity (EEG) and blood oxygenation level dependent (BOLD) signals in fMRI remains unclear.
• Simultaneous EEG-fMRI recording offers a unique opportunity to investigate brain activity.
• Post-movement beta rebound in EEG and BOLD signals in fMRI are explored for their correlation.

Purpose of the Study:

• To exploit the correlation between EEG beta rebound and fMRI BOLD signals.
• To develop a method for incorporating EEG information into fMRI analysis.
• To identify brain regions involved in EEG events using simultaneous recordings.

Main Methods:

• A variant of tensor factorization, PARAFAC, is introduced as a specific constraint for EEG-fMRI analysis.
• PARAFAC2 is employed to reveal information about fMRI BOLD signals and their time course simultaneously.
• The method allows for various constraints during parameter estimation.

Main Results:

• Extensive experiments confirm the effectiveness of the proposed PARAFAC2 method.
• The method successfully detects brain regions responsible for beta rebound.
• EEG-fMRI analysis using PARAFAC2 illustrates expected brain activities, validated against fMRI-only analysis.

Conclusions:

• The proposed method is a semi-blind decomposition technique utilizing PARAFAC2 without a predefined time course.
• This approach facilitates multi-task analysis in Brain-Computer Interface (BCI) applications.
• The findings pave the way for more sophisticated BCI development.