Correspondence of dynamic resting-state networks in source space EEG and MEG
Poster No:
1638
Submission Type:
Abstract Submission
Authors:
SungJun Cho1, Chetan Gohil1, Mats van Es1, Mark Woolrich1
Institutions:
1University of Oxford, Oxford, United Kingdom
First Author:
Co-Author(s):
Introduction:
The resting state networks (RSNs) in EEG and MEG have been pivotal for understanding the oscillatory dynamics underlying various cognitive functions and clinical conditions [1-3]. While most studies have focused on static RSNs, analysing network activities averaged over time, the dynamic aspect of these networks, especially in EEG, has received less attention. Traditional methods like microstates, sliding window correlation, and temporal independent component analysis have explored EEG-based dynamic RSNs before, but with limitations inherent to their assumptions [4-6]. The Hidden Markov Model (HMM) [7, 8] has advanced the study of dynamic RSNs in MEG, but their comparability to EEG remains underexplored. Our study addresses this gap, demonstrating that dynamic RSNs can be reliably inferred in source space EEG and showing similar group-level effects in both EEG and MEG.
Methods:
In this study, we utilized the Leipzig Study for Mind-Body-Emotion Interactions EEG dataset [9] and the Cambridge Centre for Ageing and Neuroscience MEG dataset [10]. To facilitate comparative analyses, we age-matched these data, resulting in 96 healthy subjects per dataset, each consisting of 60 young (20-35 years) and 36 older (55-80 years) adults. The sensor signals were projected onto the source space using a beamformer.
Next, we employed the Time-Delay Embedded HMM [8] to segment source reconstructed data into eight distinct dynamic RSNs, each described by a unique multivariate Gaussian model and associated oscillatory profiles. Using the inferred state activation time courses, three state-specific brain network features were extracted at the individual level: power spectral densities (PSDs), power maps, and functional connectivity (i.e., coherence) maps [11].
To contrast the ability of M/EEG in resolving the effects of interest, we investigated group-level differences between two age groups by fitting each network feature to a General Linear Model [12], with sex and head size as covariates. The statistical significance of age-related effects in network features was determined using nonparametric max-t or cluster permutation tests. The alpha significance threshold was set to 0.05.
Next, we employed the Time-Delay Embedded HMM [8] to segment source reconstructed data into eight distinct dynamic RSNs, each described by a unique multivariate Gaussian model and associated oscillatory profiles. Using the inferred state activation time courses, three state-specific brain network features were extracted at the individual level: power spectral densities (PSDs), power maps, and functional connectivity (i.e., coherence) maps [11].
To contrast the ability of M/EEG in resolving the effects of interest, we investigated group-level differences between two age groups by fitting each network feature to a General Linear Model [12], with sex and head size as covariates. The statistical significance of age-related effects in network features was determined using nonparametric max-t or cluster permutation tests. The alpha significance threshold was set to 0.05.
Results:
Our analysis revealed that dynamic RSNs display comparable subject-averaged PSDs and wide-band (1-45 Hz) power maps across both modalities in source space. Wide-band coherence maps of EEG exhibited slightly more noise compared to MEG. Overall, canonical RSNs, such as the anterior and posterior default mode networks, visual network, and sensorimotor network, were consistently identified in both modalities (Fig. 1). This aligns with the previous study that found similar spatial signatures in wide-band (3-40 Hz) power maps between EEG and MEG in sensor space [13].
Regarding age-related effects in dynamic RSNs, MEG demonstrated more pronounced age effects in PSDs, while both EEG and MEG reported similar age-related changes in narrow-band power and coherence maps. Specifically, delta (1-4 Hz) and theta (4-8 Hz) power maps indicated reduced activity in frontotemporal, parietal, and sensorimotor cortices with aging across HMM states, whereas beta (13-30 Hz) power maps showed increased activity in these regions. These age-related changes conformed to those reported in previous research [14, 15]. Age-dependent changes in coherence generally mirrored those in power but with more globally widespread activations.
Regarding age-related effects in dynamic RSNs, MEG demonstrated more pronounced age effects in PSDs, while both EEG and MEG reported similar age-related changes in narrow-band power and coherence maps. Specifically, delta (1-4 Hz) and theta (4-8 Hz) power maps indicated reduced activity in frontotemporal, parietal, and sensorimotor cortices with aging across HMM states, whereas beta (13-30 Hz) power maps showed increased activity in these regions. These age-related changes conformed to those reported in previous research [14, 15]. Age-dependent changes in coherence generally mirrored those in power but with more globally widespread activations.
Conclusions:
In summary, our findings demonstrate that dynamic RSNs can be reliably estimated from EEG, as in MEG. The brain network features derived from these RSNs robustly replicate previously reported age effects, which were comparable across modalities. These results validate the use of current methodologies and generative models in EEG analysis, corroborating the existence of canonical RSNs in EEG akin to those in MEG and fMRI. Consequently, this supports the broader application of these techniques to clinical and cognitive datasets obtained through EEG.
Lifespan Development:
Aging
Modeling and Analysis Methods:
Bayesian Modeling
Connectivity (eg. functional, effective, structural)
EEG/MEG Modeling and Analysis 1
Methods Development 2
Keywords:
Aging
Data analysis
Electroencephaolography (EEG)
Machine Learning
MEG
Modeling
Other - Resting state network
1|2Indicates the priority used for review
Provide references using author date format
[2] Sitnikova TA et al. (2018). Short timescale abnormalities in the states of spontaneous synchrony in the functional neural networks in Alzheimer’s disease. NeuroImage:Clinical, 20:128-152.
[3] Nour MM et al. (2023). Reduced coupling between offline neural replay events and default mode network activation in schizophrenia. Brain Communications, 5(2):fcad056.
[4] Custo A et al. (2017). Electroencephalographic Resting-State Networks: Source Localization of Microstates. Brain Connectivity, 7(10):671-682.
[5] Zink N et al. (2021). Resting-state EEG Dynamics Reveals Differences in Network Organization and its Fluctuation between Frequency Bands. Neuroscience, 453:43-56.
[6] Chen J et al. (2013). Dynamic Changes of ICA-Derived EEG Functional Connectivity in the Resting State. Human Brain Mapping, 34:852-868.
[7] Baker AP et al. (2014). Fast transient networks in spontaneous human brain activity. eLife, 3:e01867.
[8] Vidaurre D et al. (2018). Spontaneous cortical activity transiently organises into frequency specific phase-coupling networks. Nature Communications, 9(1):2987.
[9] Babayan A et al. (2019). A mind-brain-body dataset of MRI, EEG, cognition, emotion, and peripheral physiology in young and old adults. Scientific Data, 6:180308.
[10] Shafto MA et al. (2014). The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) study protocol: a cross-sectional, lifespan, multidisciplinary examination of healthy cognitive ageing. BMC Neurology, 14:204.
[11] Gohil C et al. (2022). Mixtures of large-scale dynamic functional brain network modes. NeuroImage, 263:119595.
[12] Quinn AJ et al. (2022). The GLM-Spectrum: A multilevel framework for spectrum analysis with covariate and confound modelling. bioRxiv:2022.11.14.516449.
[13] Coquelet N et al. (2022). Microstates and power envelope hidden Markov modeling probe bursting brain activity at different timescales. NeuroImage, 247:118850.
[14] Rossini PM et al. (2007). Clinical neurophysiology of aging brain: From normal aging to neurodegernation. Progress in Neurobiology, 83(6):375-400.
[15] Babiloni C et al. (2006). Sources of Cortical Rhythms in Adults During Physiological Aging: A Multicentric EEG Study. Human Brain Mapping, 27:162-172.