Spatial (Mis)match Between EEG and fMRI Signal Patterns Revealed by Spatio-Spectral Source-Space EEG Decomposition

Stanislav Jiříček ^1,2,3 Vlastimil Koudelka ¹ Dante Mantini ⁴ Radek Marecek ⁵ Jaroslav Hlinka ^1,2*

• ¹ National Institute of Mental Health (Czechia), Prague, Czechia
• ² Institute of Computer Science, Czech Academy of Sciences, Praha (Prague), Czechia
• ³ Department of Cybernetics, Faculty of Electrical Engineering, Czech Technical University in Prague, Prague, Prague, Czechia
• ⁴ Department of Movement Sciences, Faculty of Medicine, KU Leuven, Leuven, Belgium
• ⁵ Central European Institute of Technology (CEITEC), Brno, Olomouc, Czechia

In this work, we aimed to directly compare EEG whole-brain patterns of neural dynamics with concurrently measured fMRI BOLD data. For that purpose, we set out to derive EEG patterns based on a spatio-spectral decomposition of band-limited EEG power in the sourcereconstructed space. On a large data set of 72 subjects resting-state hdEEG-fMRI we showed that the proposed approach is reliable both in terms of the extracted patterns as well as their spatial BOLD signatures. The five most robust EEG spatio-spectral patterns not only encompass the well-known occipital alpha power dynamics, ensuring consistency with established findings, but also reveal additional patterns, uncovering new insights into brain activity. We report and interpret the most reproducible source space EEG-fMRI patterns, along with the corresponding EEG electrode space patterns better known from the literature. The EEG spatial-spectral patterns show weak, yet statistically significant spatial similarity to their fMRI BOLD signatures, particularly the patterns that show stronger temporal synchronization with BOLD. However, we didn't observe a statistically significant relation between the EEG spatio-spectral patterns and the classical fMRI BOLD resting state networks (as obtained by independent component analysis), tested as the similarity between their temporal synchronization and spatial overlap. This provides evidence that both EEG (frequency-specific) power and BOLD signal capture reproducible spatiotemporal patterns of neural dynamics. Rather than being mutually redundant, these are only partially overlapping, carrying to a large extent complementary information concerning the underlying low-frequency dynamics.