Understanding brain connectivity from EEG data by identifying systems composed of interacting sources

Area of Science:

• Neuroscience
• Biophysics
• Signal Processing

Background:

• Electroencephalography (EEG) and Magnetoencephalography (MEG) measure superimposed brain activity.
• Volume conduction in the head creates spurious interactions in EEG/MEG signals.
• Separating true source interactions from noise is crucial for understanding brain function.

Purpose of the Study:

• To develop methods for unmixing uncorrelated and correlated brain source activities.
• To distinguish true neural interactions from artifacts caused by volume conduction.
• To enhance the analysis of brain functioning using EEG/MEG data.

Main Methods:

• Developed source principal component analysis (sPCA) based on orthogonality assumptions.
• Introduced Minimum Overlap Component Analysis (MOCA) for demixing correlated sources.
• Validated methods through simulations and application to human EEG data (micro and alpha rhythms).

Main Results:

• sPCA effectively separates uncorrelated source systems.
• MOCA successfully unmixes correlated sources within identified systems.
• The combined approach accurately distinguishes true neural interactions in simulated and real EEG data.

Conclusions:

• sPCA and MOCA provide robust tools for source separation in EEG/MEG.
• These methods improve the accuracy of brain activity analysis by mitigating volume conduction effects.
• The findings facilitate a deeper understanding of neural interference mechanisms.