Vigilance declines following sleep deprivation are associated with two previously identified dynamic connectivity states

Highlights

• 32 individuals underwent task-free fMRI twice: while well-rested and sleep deprived.

• Clustering of dynamic connectivity matrices replicated centroids found previously.

• The profile of time spent in each connectivity state changed with sleep deprivation.

• Changes in time spent in arousal-related states correlated with declines in vigilance.

Abstract

Robustly linking dynamic functional connectivity (DFC) states to behaviour is important for establishing the utility of the method as a functional measurement. We previously used a sliding window approach to identify two dynamic connectivity states (DCS) related to vigilance. A new sample of 32 healthy participants underwent two sets of task-free functional magnetic resonance imaging (fMRI) scans, once in a well-rested state and once after a single night of total sleep deprivation. Using a temporal difference method, DFC and clustering analysis on the task-free fMRI data revealed five centroids that were highly correlated with those found in previous work. In particular, two of these states were associated with high and low arousal respectively. Individual differences in vulnerability to sleep deprivation were measured by assessing state-related changes in Psychomotor Vigilance Test (PVT) performance. Changes in the duration spent in each of the arousal states from the well-rested to the sleep-deprived condition correlated with declines in PVT performance. The reproducibility of DFC measures and their association with vigilance highlight their utility in serving as a neuroimaging method with behavioural relevance. (178 words).

Introduction

Dynamic functional connectivity (DFC) analysis of neuroimaging data is increasingly being used to study how inter-region connectivity strength and network configurations evolve over time (Hutchison et al., 2013). One common approach in DFC analysis is to search within a connectivity time series (Allen et al., 2014) to identify recurring patterns known as dynamic connectivity states (DCSs). There is now some evidence that DCSs contain information that is of behavioural significance. For instance, on a coarse level, certain DCS have been associated with wakefulness and sleep (Haimovici et al., 2017; Damaraju et al., 2018), motivating a more detailed search for states related to other cognitive domains, such as sustained attention, mind wandering, or even spontaneous thoughts (Kucyi et al., 2018).

In the short time since the first reports of time-varying connectivity, a plethora of different approaches have been proposed to derive DFC estimates, each with their own theoretical underpinnings. These approaches vary along three major dimensions: 1) the transformations applied to the data, 2) the function used to quantify relationships between windows within the time series, and 3) the weighting vectors applied to the relational computation (Thompson and Fransson, 2018). Within each type of analysis are parameters that can be tuned (e.g. window size for analyses involving moving averages, component selection for ICA approaches). This heterogeneity in methods complicates the interpretation and comparison of findings, and has made it challenging to link particular DCS to specific cognitive or behavioural states.

Several reports have been published on the reliability of DFC estimates, reaching the general conclusion that reliability is good for summary statistics (e.g. average connectivity, percentage of occurrence) and connectivity features, but relatively lower for derived measures such as dwell time and transitions (Abrol et al., 2017; Choe et al., 2017; Smith et al., 2018). However, it is typically these latter measures that are used in a search for DFC-behaviour relationships. It thus follows that a system employed to test the robustness of DFC-behaviour associations must be anchored by a highly reliable behavioural phenomenon.

One such phenomenon is the large declines in sustained attention that follow acute sleep deprivation (SD) (Lim and Dinges, 2010). Individual differences in these impairments are stable over time (Leproult et al., 2003; Van Dongen et al., 2004), and this trait-like nature makes them especially amendable to reproducibility studies. State-related shifts in BOLD activation in frontoparietal regions are reproducible across two nights of total SD (Lim et al., 2007). State-related shifts of task activation and corresponding shifts in sustained attention are also reproducible within different conditions of the same experiment (Chee and Tan, 2010). These findings make a night of total sleep deprivation an attractive test bed to study how DFC measures uncovered using one method generalize across another analysis methodology.

Using static, or time-averaged connectivity analysis, it has been shown that the brain connectome is less integrated and segregated in SD (Yeo et al., 2015), and that anti-correlations are particularly relevant markers of state differences (Samann et al., 2010; De Havas et al., 2012). Using DFC analysis, Xu et al. (2018) showed that large shifts in the proportions of dwell times and the transition probability matrix occur in resting-state data after 36 h of total SD. Separately, two prior DFC studies have reported DCSs associated with vigilance (the high arousal state (HAS) and low arousal state (LAS)), showing that these DCSs in the sleep-deprived state are associated with temporal fluctuations in vigilance at rest and in an auditory vigilance task condition (Wang et al., 2016), and can be used to predict vulnerability to SD while individuals are still in a well-rested state (Patanaik et al., 2018).

In the current study, we exploited SD as a tool to test the robustness of the HAS/LAS behavioural associations across datasets and analysis methods (i.e. in comparison with our previous published findings). Specifically, we sought to elaborate on our previous findings by directly investigating SD-related individual differences in DCSs, and vigilant attention. To achieve this, we collected task-free fMRI data at baseline and after 24 h of total sleep deprivation in a group of 32 healthy young adults. Our primary hypothesis was that following SD, we would observe reductions in a composite measure calculated from proportions of two DCSs previously shown to index high and low arousal states (Patanaik et al., 2018), and that this decrement would correlate with state-related shifts in vigilance. A third state previously associated with trait mindfulness (Lim et al., 2018), and two other unnamed (but reproducible) states were also tested to demonstrate the specificity of behavioural associations related to the aforementioned high and low arousal states. Critically, we used a different method of DFC computation (multiplication of temporal differences) than in our original reports on DCS relationship with arousal for all these analyses. Finally, we tested two other relevant measures that are affected by arousal – global signal variability and head motion – to assess their independent contribution to predicting state-shifts in vigilance.