Frequency analysis of EEG microstate sequences in wakefulness and NREM sleep

Abstract

The majority of EEG microstate analyses concern wakefulness, and the existing sleep studies have focused on changes in spatial microstate properties and on microstate transitions between adjacent time points, the shortest available time scale. We present a more extensive time series analysis of unsmoothed EEG microstate sequences in wakefulness and non-REM sleep stages across many time scales. Very short time scales are assessed with Markov tests, intermediate time scales by the entropy rate and long time scales by a spectral analysis which identifies characteristic microstate frequencies.During the descent from wakefulness to sleep stage N3, we find that the increasing mean microstate duration is a gradual phenomenon explained by a continuous slowing of microstate dynamics as described by the relaxation time of the transition probability matrix. The finite entropy rate, which considers longer microstate histories, shows that microstate sequences become more predictable (less random) with decreasing vigilance level. Accordingly, the Markov property is absent in wakefulness but in sleep stage N3, 10/19 subjects have remarkably simple microstate sequences which are compatible with a second-order Markov process.A spectral microstate analysis is performed by comparing the time-lagged mutual information coefficients of microstate sequences with the autocorrelation function of the underlying EEG. We find periodic microstate behavior in all vigilance states, linked to alpha frequencies in wakefulness, theta activity in N1, sleep spindle frequencies in N2, and in the delta frequency band in N3.In summary, we show that EEG microstates are a dynamic phenomenon with oscillatory properties that slow down in sleep and are coupled to specific EEG frequencies across several sleep stages.

Complexity measures for EEG microstate sequences - concepts and algorithms

“…Entropy rate estimation for microstate sequence analysis was introduced in (von Wegner et al, 2018a), and we evaluated its changes during different types of cognitive effort in (Jia et al, 2021), and for NREM sleep stages in (Wiemers et al, 2022). We interpreted entropy rate in terms of sequence predictability in (von Wegner et al, 2018a) and (Jia et al, 2021), and as a complexity measure in (Wiemers et al, 2022).…”

“…In the second part of the results section, we evaluate the four metrics on EEG microstate sequences in wakefulness and non-REM (NREM) sleep, a dataset we have previously analyzed with other microstate analysis tools (Brodbeck et al, 2012;Wiemers et al, 2022). In an attempt to identify which time series features the different complexity metrics actually 'see', we use first-order Markov surrogate data to represent exactly that amount of information that is captured by the transition probability matrix, an approach that is often used to report microstate data (Lehmann et al, 2005).…”

“…We analyzed EEG recordings from n=19 healthy subjects. The EEG dataset is a subset of the dataset analyzed in (Brodbeck et al, 2012;Wiemers et al, 2022) and only includes those subjects for whom sleep stage N3 data was available. Briefly, EEG data from simultaneous EEG-fMRI recordings were corrected for scanner and cardioballistic artefacts and downsampled to 250 Hz as described in (Brodbeck et al, 2012).…”

“…No further smoothing methods were applied. Microstate maps and further properties of this dataset are detailed in (Wiemers et al, 2022).…”

“…Entropy rate and LZC have so far been discussed separately in the microstate literature (von Wegner et al, 2018a;Jia et al, 2021;Tait et al, 2020;Artoni et al, 2022;Wiemers et al, 2022). The equivalence of both metrics means that LZC values as those given in (Tait et al, 2020) and entropy values we published in different contexts (von Wegner and Laufs, 2018;Jia et al, 2021;Wiemers et al, 2022) can be re-scaled and compared directly. Unfortunately, this is not possible for LZC values computed by more recent versions of the LZ algorithm, for example the complexity values given in (Artoni et al, 2022).…”