Differences in MEG and EEG power-law scaling explained by a coupling between spatial coherence and frequency: a simulation study

TL;DR

This simulation study explores how the spatial and frequency structure of neural sources influences the power-law spectral differences observed between EEG and MEG signals, providing insights into their underlying neurophysiological mechanisms.

Contribution

It introduces a simulation framework demonstrating how source size and spatial coherence affect EEG and MEG spectral scaling, explaining observed differences.

Findings

Space/frequency structure causes EEG/MEG spectral differences compatible with real data.

Below a certain spatial scale, EEG and MEG spectra become indistinguishable.

Results suggest a resolution limit for EEG and MEG in detecting neural scale differences.

Abstract

Electrophysiological signals (electroencephalography, EEG, and magnetoencephalography , MEG), as many natural processes, exhibit scale-invariance properties resulting in a power-law (1/f) spectrum. Interestingly, EEG and MEG differ in their slopes, which could be explained by several mechanisms, including non-resistive properties of tissues. Our goal in the present study is to estimate the impact of space/frequency structure of source signals as a putative mechanism to explain spectral scaling properties of neuroimaging signals. We performed simulations based on the summed contribution of cortical patches with different sizes (ranging from 0.4 to 104.2 cm 2). Small patches were attributed signals of high frequencies, whereas large patches were associated with signals of low frequencies, on a logarithmic scale. The tested parameters included i) the space/frequency structure (range of patch sizes and frequencies) and ii) the amplitude factor c parametrizing the spatial scale ratios. We found that the space/frequency structure may cause differences between EEG and MEG scale-free spectra that are compatible with real data findings reported in previous studies. We also found that below a certain spatial scale, there were no more differences between EEG and MEG, suggesting a limit for the resolution of both methods. Our work provides an explanation of experimental findings. This does not rule out other mechanisms for differences between EEG and MEG, but suggests an important role of spatio-temporal structure of neural dynamics. This can help the analysis and interpretation of power-law measures in EEG and MEG, and we believe our results can also impact computational modeling of brain dynamics, where different local connectivity structures could be used at different frequencies.