Comparing MEG and high-density EEG for intrinsic functional connectivity mapping
-
1
Free University of Brussels, Belgium
Several resting-state networks (RSNs) have been identified using magnetoencephalography (MEG).[1,2] More recently, high-density electroencephalography (hdEEG) has become an additional tool in the armamentarium of RSN analysis.[3,4] However, the hdEEG literature on RSNs currently lacks two major aspects that have been investigated with MEG: (1) the global assessment of the whole brain connectome including both intra- and cross-RSNs interactions, and (ii) the description of the dynamic modulation of RSNs and connectome. In this study, we compare the resting-state functional connectivity (rsFC) patterns, both in their static and dynamic aspects, directly using simultaneous MEG and hdEEG resting-state data. We also assess the effect of hdEEG forward modeling on rsFC estimation.
Twenty-four healthy subjects (age: 26±4.3 years, mean±std, 10 females) underwent simultaneous MEG-hdEEG resting-state recordings (5 minutes eyes open; MEG: Vectorview (for 14 subjects) & Triux (for 10 subjects), Elekta Oy; EEG: 256-channel MicroCel Geodesic Sensor Net, EGI). External MEG noise was reduced using signal-space separation whereas electrode noise was reduced using bad channel detection and replacement. MEG forward modelling, which is relatively insensitive to head conductivity, was based on the one-layer Boundary Element Method (BEM). For hdEEG, we used either three-layer BEM or five-compartment Finite Element Method (FEM). All forward models relied on individually-segmented anatomical images and default tissue compartmentation and conductivity (MNE-C & Fieldtrip Simbio).[5] Source projection in the alpha (α: 8-13 Hz) and beta (β: 12-30 Hz) frequency bands was performed using Minimum Norm Estimation. We estimated rsFC using source envelope correlation with orthogonalization for leakage correction (static: computed over 5 min, dynamic: computed within 10-s long sliding windows, 5 s overlap). Although our main focus was on the connectome, we started by a quality control analysis of seed-based static rsFC maps of five well-known RSNs in their respective frequency band: 3 classical primary RSNs (sensorimotor, auditory and visual primary networks), the default-mode and the fronto-parietal networks. We then turned to the connectome procedure, where rsFC was computed among the signals of 82 cortical parcels covering the whole brain. The dynamic rsFC data were further classified into k different rsFC states using k-means clustering (with correlation as clusterization index). Each state was characterized by a rsFC connectome pattern and a binary time series of appearance. The topographical similarity of static rsFC (seed-based maps or connectome) across the three modalities (MEG, BEM3- and FEM5-hdEEG) was quantified using spatial Pearson correlations (one-tailed parametric tests). The regional differences in connectomes were further localized using non-parametric permutation testing (see [6] for additional information). Finally, we sought for temporally paired rsFC states between MEG on the one hand and BEM3-hdEEG and FEM5-hdEEG on the other hand using Spearman correlation of their binary time series (one-tailed parametric tests). The topographical similarity of significant state pairs was also assessed as described above. All tests were performed at significance level p<0.05.
The three classical primary RSNs were recognizable within each modality, and the corresponding cross-modality similarities were all significant. For the DMN, the typical antero-posterior interactions were observable with MEG and FEM5-hdEEG but not with BEM3-hdEEG. The topography of fronto-parietal network maps disclosed a notable dissimilarity between MEG (disclosing the intra-hemispheric fronto-parietal connections) and both hdEEG modalities (showing the inter-hemispheric frontal connectivity). Accordingly, at the level of the static connectome, hdEEG - independently of the forward model - showed higher regional rsFC for nodes in the frontal regions, and MEG, higher regional rsFC within the parieto-occipital regions. Those patterns are reminiscent of the topography of brain-sensor distance, as posterior MEG sensors were closer to the scalp and frontal MEG sensors were further away. After regressing out this parameter, no significant residual contrast subsisted between MEG and hdEEG regional rsFC. Finally, at the level of rsFC state dynamics, we found only three temporal state pairs in the α band (one MEG - BEM3-hdEEG and two MEG - FEM5-hdEEG) and none in the β band.
In conclusion, our study demonstrated that intrinsic rsFC patterns reconstructed from MEG and scalp hdEEG are similar at the level of static rsFC, provided that brain-sensor distance is taken into account. This result concurs with the hypothesis of [4] that rsFC and RSNs are generated by cortical areas large enough to cover both sulci and gyri and thus mitigate the differences in hdEEG and MEG sensitivity to precise source location and orientation. We also showed that the associated rsFC state dynamics is highly discordant between the two modalities. This discrepancy between MEG and hdEEG suggests that short-time rsFC may have higher sensitivity to different generators of the spontaneous neuromagnetic and neuroelectric fields. Finally, we highlighted that the choice of hdEEG forward models has only a major impact on the DMN as the FEM5 forward model was able to detect the characteristic antero-posterior rsFC of the DMN while mapping the other RSNs was relatively insensitive to the choice of the forward model.
References
[1] Brookes et al. PNAS, 2011
[2] Hipp et al. Nat Neurosci, 2012
[3] Liu et al. HBM, 2017
[4] Siems et al. Neuroimage, 2016
[5] Vorwerk et al. Biomed Eng Online, 2018
[6] Wens et al. HBM, 2015
Keywords:
magnetoencephalagraphy,
Electroencepahlography,
resting-state network (RSN),
forward model,
Connectome Analysis,
resting state,
dynamic
Conference:
13th National Congress of the Belgian Society for Neuroscience , Brussels, Belgium, 24 May - 24 May, 2019.
Presentation Type:
Poster presentation
Topic:
Behavioral/Systems Neuroscience
Citation:
Coquelet
N,
De Tiège
X,
Destoky
F,
Roshchupkina
L,
Bourguignon
M,
Goldman
S,
Peigneux
P and
Wens
V
(2019). Comparing MEG and high-density EEG for intrinsic functional connectivity mapping.
Front. Neurosci.
Conference Abstract:
13th National Congress of the Belgian Society for Neuroscience .
doi: 10.3389/conf.fnins.2019.96.00026
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
30 Apr 2019;
Published Online:
27 Sep 2019.
*
Correspondence:
Mr. Nicolas Coquelet, Free University of Brussels, Brussels, Belgium, nico.coquelet@gmail.com