Realistic Subject-Specific Simulation of Resting State Scalp EEG Based on Physiological Model

Area of Science:

• Computational Neuroscience
• Neuroimaging
• Brain-Computer Interfaces

Background:

• Electroencephalography (EEG) is crucial for understanding brain function and disease, but cellular origins of scalp signals remain unclear in humans.
• Current limitations restrict cellular-level EEG recordings to animal models, hindering human brain research.

Purpose of the Study:

• To simulate individual-specific spatiotemporal features of resting-state (RS) EEG in humans.
• To assess the similarity between real and computationally generated EEG signals.
• To advance data augmentation for AI and brain-computer interfaces.

Main Methods:

• Developed a physiologically grounded whole-brain computational model with detailed neuronal subtypes and connectivity.
• Simulated interregional cortical circuitry activity to generate realistic individual RS EEG.
• Utilized high-definition EEG and source localization to analyze alpha and beta-gamma rhythms.
• Created a similarity index based on cross-correlation analysis to evaluate simulated EEG realism.

Main Results:

• Successfully generated realistic individual RS EEG rhythms using the computational model.
• Identified specific neural mechanisms: somatostatin-pyramidal loop for alpha oscillations (posterior) and parvalbumin-interneuron excitability for beta-gamma oscillations (anterior).
• Demonstrated high similarity between simulated and real EEG signals.

Conclusions:

• The computational model accurately reproduces individual-specific resting-state EEG rhythms.
• This simulation capability offers a significant advancement for data augmentation in neuroscience research.
• Enables new possibilities for brain-computer interfaces and artificial intelligence training using realistic EEG data.