Exploring the Interplay of EEG and BOLD Signals: Insights from Neural Mass Modeling

1674 

Abstract Submission 

Shih-Cheng Chien¹, Stanislav Jiříček², Thomas Knösche³, Jaroslav Hlinka², Helmut Schmidt¹

¹Institute of Computer Science, The Czech Academy of Sciences, Prague, Prague, ²Institute of Computer Science of the Czech Academy of Sciences, Prague, Czech Republic, ³Max Planck Institute, Leipzig, Saxony

Shih-Cheng Chien  
Institute of Computer Science, The Czech Academy of Sciences
Prague, Prague

Stanislav Jiříček  
Institute of Computer Science of the Czech Academy of Sciences
Prague, Czech Republic

Thomas Knösche  
Max Planck Institute
Leipzig, Saxony

Jaroslav Hlinka  
Institute of Computer Science of the Czech Academy of Sciences
Prague, Czech Republic

Helmut Schmidt  
Institute of Computer Science, The Czech Academy of Sciences
Prague, Prague

Studies investigating the relationship between electrophysiology (EEG) rhythms and blood-oxygen-level-dependent (BOLD) activity have revealed intriguing correlations, such as the negative correlation between EEG alpha power and BOLD signal in occipital areas during eyes-closed resting states. Conversely, positive correlations have been observed in various other brain regions [1]. Additionally, studies have identified positive correlations between EEG gamma power and the BOLD signal during sensory stimulation [2]. A possible interpretation for these diverse findings may lie in the involvement of distinct types of inhibitory neurons, known for their differential effects on EEG rhythm generation [3] and varied neural-vascular coupling properties [4].

This study employs a cortical column model to simulate EEG and BOLD signals, encompassing excitatory (E) and inhibitory (PV, SOM, and VIP) populations across cortical layers (L2/3, L4, and L5/6). Various conditions are simulated, with the model initially driven by constant inputs to induce either PV-dominant or SOM-dominant states [5]. Subsequently, a combination of pink noises serves as thalamic input (to E and PV populations) and cortico-cortical input (to E populations). The EEG signals are approximated by the weighted sum of simulated excitatory and inhibitory postsynaptic potentials (EPSPs and IPSPs) of the E populations. BOLD signals are simulated using the Balloon-Windkessel model [6], where the input is the sum of simulated EPSPs of the E, PV, SOM, and VIP populations. Simulated EEGs and BOLDs undergo typical analysis procedures employed in investigating the EEG-BOLD relationship in experimental studies.

In the PV-dominant state, our simulations reveal a positive correlation between the EEG and BOLD signals, denoted by a significant correlation coefficient. Conversely, the simulated EEG and BOLD signals exhibit a negative correlation under the SOM-dominant state. These findings highlight the influence of inhibitory neuron dominance on the observed correlation patterns between EEG and BOLD signals.

This preliminary exploration delves into the potential factors contributing to the observed (anti-)correlation between EEG and BOLD signals in EEG-fMRI experiments. We identify instances that give rise to the observed anti/correlation phenomena by systematically examining model configurations across the parameter space. These configurations are linked to dynamic factors, such as variations in input sets, and static factors, including intra-column connectivity and the ratio of PV/SOM cell counts. The nuanced interplay of these factors provides valuable insights for future investigations into the underlying dynamic columnar states in simultaneous EEG/fMRI studies.

EEG/MEG Modeling and Analysis ¹

fMRI Connectivity and Network Modeling

Task-Independent and Resting-State Analysis ²

Cerebral Blood Flow

Computational Neuroscience

Cortex

Cortical Columns

Cortical Layers

Electroencephaolography (EEG)

FUNCTIONAL MRI

Modeling