MEG cortical microstates: spatiotemporal characteristics, dynamic functional connectivity and stimulus-evoked responses

Abstract

EEG microstate analysis is a useful approach for studying brain states - nicknamed `atoms of thought' - and their fast transitions in healthy cognition and disease. A key limitation of conventional microstate analysis is that it must be performed at the sensor level, and therefore gives limited anatomical insight into the cortical mechanisms underpinning these states. In this study, we generalise the microstate methodology to be applicable to source-reconstructed electrophysiological data. Using simulations of a neural-mass network model, we first established the validity and robustness of the proposed method. Using MEG resting-state data, we uncovered ten microstates with distinct spatial distributions of cortical activation. Multivariate pattern analysis demonstrated that source-level MEG microstates were associated with distinct functional connectivity patterns. Using a passive auditory paradigm, we further demonstrated that the occurrence probability of MEG microstates were altered by evoked auditory responses, exhibiting a hyperactivity of the microstate including the auditory cortex. Our results support the use of MEG source-level microstates as a data-driven method for investigating brain dynamic activity and connectivity at the millisecond scale.

+microstate: A MATLAB toolbox for brain microstate analysis in sensor and cortical EEG/MEG

“…However, a key limitation of the EEG microstate pipeline is that it is unsuitable for application to MEG or source-reconstructed M/EEG data (Tait and Zhang, 2021). Since MEG has higher spatial resolution than EEG (Boly et al, 2015), and source-reconstruction allows for anatomical interpretation of the electrophysiological data on the cortical level (He et al, 2018), generalization of the microstate pipeline to these modalities is crucial for advancement of understanding the neural mechanisms underpinning brain microstates.…”

“…Since MEG has higher spatial resolution than EEG (Boly et al, 2015), and source-reconstruction allows for anatomical interpretation of the electrophysiological data on the cortical level (He et al, 2018), generalization of the microstate pipeline to these modalities is crucial for advancement of understanding the neural mechanisms underpinning brain microstates. Recently, we presented such a generalization of the microstate k -means algorithm and applied this algorithm to uncover source-space resting-state microstates and their associations with auditory stimulation (Tait and Zhang, 2021). A number of further advancements were also presented, including validation that microstates were associated with distinct patterns of cortical synchrony and a pipeline to simulate M/EEG sensor- or source-space data with known ground truth microstate maps and microstate sequences (Tait and Zhang, 2021).…”

“…Recently, we presented such a generalization of the microstate k -means algorithm and applied this algorithm to uncover source-space resting-state microstates and their associations with auditory stimulation (Tait and Zhang, 2021). A number of further advancements were also presented, including validation that microstates were associated with distinct patterns of cortical synchrony and a pipeline to simulate M/EEG sensor- or source-space data with known ground truth microstate maps and microstate sequences (Tait and Zhang, 2021).…”

“…Here, we present +microstate, an open-source MATLAB toolbox for multi-modal analysis of microstates in M/EEG sensor- and source-space using the generalized k -means algorithm presented by Tait and Zhang (2021). Several open-source toolboxes for sensor-space EEG microstate analysis currently exist such as Cartool, a plugin for the BrainVision Analyzer, and a plugin for EEGLAB (Michel and Koenig, 2018), but +microstate is the only toolbox available for MEG or source-reconstructed microstate analysis.…”

“…While the focus is on k -means, +microstate includes options to perform Hidden Markov Modelling (HMM; Baker et al (2014); Vidaurre et al (2018)), principal component analysis (PCA), and independent component analysis (ICA; Hyvarinen (1999)) for brain-state estimation using external packages (see subsubsection 2.1.1) in the same framework as our microstate pipeline so that brain-state dynamics using different methodologies can be directly compared and contrasted. Furthermore, +microstate includes functions to calculate microstate-segmented functional connectivity (Hatz et al, 2015, 2016; Tait and Zhang, 2021) which is potentially useful as a tool for studying dynamic functional connectivity beyond the arbitrary sliding window (Tait and Zhang, 2021), as well as perform simulations of data from ground truth microstate sequences (Tait and Zhang, 2021), which is useful for assessing clustering methodologies against a ground truth (Tait and Zhang, 2021). Novel advanced statistical techniques for task-based analysis are additionally included in the toolbox, including cluster-permutation (Maris and Oostenveld, 2007) extensions to topographic ANOVAs (TANOVAs) (Murray et al, 2008) and χ 2 statistics of microstate probabilities in response to stimuli (Tait and Zhang, 2021).…”