Uncovering a stability signature of brain dynamics associated with meditation experience using massive time-series feature extraction

Abstract

Previous research has examined resting electroencephalographic (EEG) data to explore brain activity related to meditation. However, this research has mostly examined power in different frequency bands. Here we compared >7000 time-series features of the EEG signal to comprehensively characterize brain activity differences between meditators and non-meditators. Eyes-closed resting-state EEG data from 49 meditators and 46 non-meditators was decomposed into the top eight principal components (PCs). We extracted 7,381 time-series features from each PC and each participant and used them to train classification algorithms to identify meditators. Highly differentiating individual features from successful classifiers were analysed in detail. Only the third PC showed above-chance classification accuracy (67%, pFDR = 0.007), for which 405 features significantly distinguished meditators (all pFDR < 0.05). Top-performing features indicated that meditators exhibited more consistent statistical properties across shorter subsegments of their EEG time-series (higher stationarity) and displayed an altered distributional shape of values about the mean. By contrast, classifiers trained with traditional band-power measures did not distinguish the groups (pFDR > 0.05). Our novel analysis approach suggests the key signatures of meditators' brain activity are higher temporal stability and a distribution of time-series values suggestive of longer, larger, or more frequent non-outlying voltage deviations from the mean within the third PC of their EEG data. The higher temporal stability observed in this EEG component might underpin the higher attentional stability associated with meditation. The novel time-series properties identified here have considerable potential for future exploration in meditation research and the analysis of neural dynamics more broadly.

Experienced meditators show greater forward travelling cortical alpha wave strengths

“…Furthermore, the study using EEG and modelling showed that the enhanced thalamocortical functional connectivity was associated with increased dynamic stability of the EEG activity (Saggar et al, 2015). Applying a comprehensive data-driven approach to analyse the resting EEG data reported in the current study, we also found that meditators had higher dynamic stability of their EEG activity compared to the non-meditators (Bailey et al, 2024).…”

“…The first study included a total of 34 experienced meditators and 36 healthy control non-meditators, and the second study included a total of 39 meditators and 36 healthy control non-meditators. Both studies included the recording of eyes-closed resting EEG data (Bailey et al, 2024; McQueen et al, 2023) and a Go/No-go task that presented Go and No-go trials with equal probability, with no more than three of each trial type presented consecutively (Bailey, Freedman, et al, 2019; Bailey et al, 2023b; Bailey, Raj, et al, 2019). However, due to time constraints within the recording sessions, not all participants provided both types of recordings, and some recordings were excluded due to artifact contamination.…”

“…This instruction was provided to enable characterisation of potential trait differences in resting EEG activity, rather than differences related to the meditation state. More details on these datasets can be found in Bailey et al (2024) for the resting data and Bailey, Freedman, et al (2019), Bailey, Raj, et al (2019), and Bailey et al (2023b) for the Go/No- go data. Additionally, 29 meditators and 29 non-meditators from the first study completed a modified Sternberg working memory task (Bailey et al, 2020).…”

“…More details on these datasets can be found in Bailey et al (2024) for the resting data and , Bailey et al (2023b) for the Go/Nogo data. Additionally, 29 meditators and 29 non-meditators from the first study completed a modified Sternberg working memory task (Bailey et al, 2020).…”

Canonical time-series features for characterizing biologically informative dynamical patterns in fMRI

“…For studies that handselect a single feature of interest (like a metric of autocorrelation timescale) [4,5], the large range of alternative features are left untested, leaving open the possibility that alternative statistics could be more informative and interpretable. Applications that have systematically compared the performance across a large feature library, such as hctsa [3,8,17,6,18,19], have involved substantial computational expense and statistical care in dealing with high-dimensional feature spaces (often requiring correction across thousands of features, limiting the power to detect signals in smaller samples). fMRI time series are short, noisy, and-particularly in the restingstate regime where dynamics may be approximated as close to a steady-state-are generally well-suited to being characterized by linear time-series analysis methods [20,21].…”