Transient spectral events in resting state MEG predict individual time-frequency task responses

Abstract

Even in response to apparently simple tasks such as hand moving, human brain activity shows remarkable inter-subject variability. Presumably, this variability reflects genuine behavioural or functional variability. Recently, spatial variability of resting-state features in fMRI -specifically connectivity -has been shown to explain (spatial) taskresponse variability. Such a link, however, is still missing for M/EEG data and its spectrally rich structure. At the same time, it has recently been shown that task responses in M/EEG can be well represented using transient spectral events bursting at fast time scales. Here, weshow that individual differences in the spatio-spectral structure of M/EEG task responses, can, to a reasonable degree, be predicted from individual differences in transient spectral events identified at rest. In a MEG dataset of diverse task conditions (including motor responses, working memory and language comprehension tasks) and resting-state sessions for each subject (n = 89), we used Hidden-Markov-Modelling to identify transient spectral events as a feature set to learn the mapping of space-time-frequency content from rest to task. Resulting trial-averaged, subject-specific task-response predictions were then compared with the actual task responses in left-out subjects. All task conditions were predicted significantly above chance. Furthermore, we observed a systematic relationship between genetic similarity (e.g. unrelated subjects vs. twins) and predictability. These findings support the idea that subject-specific transient spectral events in resting-state neural activity are linked to, and predictive of, subject-specific trial-averaged task responses in a wide range of experimental conditions.

Spontaneous network activity <35 Hz accounts for variability in stimulus-induced gamma responses

“…So far, there is only one comparable piece of work from Becker et al, who likewise combined MEG and HMMs to predict electrophysiological responses to visual stimuli and own movements from resting-state activity (Becker et al, 2018). The current study differs from this paper in several ways.…”

“…Other features such as frequency or latency were not assessed. And, unlike previous fMRI studies (Cole et al, 2016;Tavor et al, 2016) and the MEG study by Becker et al (Becker et al, 2018), it did not probe the predictive power of resting-state activity by generating out-of-sample predictions.…”

“…According to our hypothesis, these inter-individual differences become apparent even in the absence of gamma-inducing stimuli, implying that resting-state activity can predict gamma responses. This idea is based on functional magnetic resonance imaging (fMRI) (Smith et al, 2009;Cole et al, 2014Cole et al, , 2016Tavor et al, 2016) and one recent MEG study (Becker et al, 2018), which demonstrated that resting-state network activity predicts inter-individual differences in task-related brain activity. Rest-task cross-frequency relationships affecting gamma oscillations, however, have not been investigated so far.…”