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

Electroencephalography (EEG) recordings are widely used in neuroscience to identify healthy individual brain rhythms and to detect alterations associated with various brain diseases. However, understanding the cellular origins of scalp EEG signals and their spatiotemporal changes during the resting state (RS) in humans remains challenging, as cellular-level recordings are typically restricted to animal models. The objective of this study was to simulate individual-specific spatiotemporal features of RS EEG and measure the degree of similarity between real and simulated EEG. Using a physiologically grounded whole-brain computational model (based on known neuronal subtypes and their structural and functional connectivity) that simulates interregional cortical circuitry activity, realistic individual EEG recordings during RS of three healthy subjects were created. The model included interconnected neural mass modules simulating activities of different neuronal subtypes, including pyramidal cells and four types of GABAergic interneurons. High-definition EEG and source localization were used to delineate the cortical extent of alpha and beta-gamma rhythms. To evaluate the realism of the simulated EEG, we developed a similarity index based on cross-correlation analysis in the frequency domain across various bipolar channels respecting standard longitudinal montage. Alpha oscillations were produced by strengthening the somatostatin-pyramidal loop in posterior regions, while beta-gamma oscillations were generated by increasing the excitability of parvalbumin-interneurons on pyramidal neurons in anterior regions. The generation of realistic individual RS EEG rhythms represents a significant advance for research fields requiring data augmentation, including brain-computer interfaces and artificial intelligence training.