Elsevier

NeuroImage

Volume 202, 15 November 2019, 116129
NeuroImage

Imaging the spontaneous flow of thought: Distinct periods of cognition contribute to dynamic functional connectivity during rest

https://doi.org/10.1016/j.neuroimage.2019.116129Get rights and content

Highlights

  • Covert self-driven cognition is a key contributor to dynamic FC at rest.

  • Several behaviorally-relevant FC configurations occur during a single rest scan.

  • Propose a way to map dynamic FC onto cognitive states in an open-ended fashion.

  • Inferred mental states agree with prior reports of common resting mental activities.

Abstract

Brain functional connectivity (FC) changes have been measured across seconds using fMRI. This is true for both rest and task scenarios. Moreover, it is well accepted that task engagement alters FC, and that dynamic estimates of FC during and before task events can help predict their nature and performance. Yet, when it comes to dynamic FC (dFC) during rest, there is no consensus about its origin or significance. Some argue that rest dFC reflects fluctuations in on-going cognition, or is a manifestation of intrinsic brain maintenance mechanisms, which could have predictive clinical value. Conversely, others have concluded that rest dFC is mostly the result of sampling variability, head motion or fluctuating sleep states. Here, we present novel analyses suggesting that rest dFC is influenced by short periods of spontaneous cognitive-task-like processes, and that the cognitive nature of such mental processes can be inferred blindly from the data. As such, several different behaviorally relevant whole-brain FC configurations may occur during a single rest scan even when subjects were continuously awake and displayed minimal motion. In addition, using low dimensional embeddings as visualization aids, we show how FC states—commonly used to summarize and interpret resting dFC—can accurately and robustly reveal periods of externally imposed tasks; however, they may be less effective in capturing periods of distinct cognition during rest.

Introduction

As the field of functional Magnetic Resonance Imaging (fMRI) moves away from group-level reports and becomes increasingly interested in the types of single-subject evaluations that are necessary for clinical diagnosis and prognosis; there is an increasing focus on examining how brain states change over time rather than treating each scan as a static snapshot of a person’s brain function (Hutchison et al., 2013a; Preti et al., 2017; Saggar et al., 2018). For example, individual scans today are often described in terms of a limited set of recurring, short-duration (tens of seconds), whole-brain FC configurations named FC states (Allen et al., 2014; Damaraju et al., 2014; Gonzalez-Castillo et al., 2015). Metrics describing their dwell times, ordering and frequency of transitions can then be used to quantify different aspects of empirically observed dFC (Calhoun et al., 2014). Many questions remain both about the etiology of empirically observed systems-level FC dynamics; as well as regarding the ability of current models to accurately capture behavioral, cognitive and clinically relevant dynamic phenomena.

One well accepted principle is that task performance can modulate dFC across the brain (Gonzalez-Castillo and Bandettini, 2018), even if the brain never abandons an overall small-world configuration (Di et al., 2013). As humans perceive external stimuli and engage with cognitive tasks, patterns of communication across brain networks reshape (Cole et al., 2014), modularity decreases (Kitzbichler et al., 2011), communication hubs relocate (Leske et al., 2015), and overall variability of FC estimates drops (Elton and Gao, 2015). FC modulations induced by tasks are sufficiently profound as to allow prediction of subsequent perceptual outcomes such as response time (Thompson et al., 2013) and pain levels (Ploner et al., 2010). Furthermore, FC estimates computed from only seconds long portions of data can be used to classify scans according to tasks (Shirer et al., 2012) and to segment multi-task scans into task-homogenous periods (Gonzalez-Castillo et al., 2015).

Nonetheless, despite empirical reports of dFC in resting humans (Allen et al., 2014), macaques (Hutchison et al., 2013b) and rodents (Keilholz et al., 2013), a similar level of consensus does not exist regarding the significance of dFC phenomena while at rest—with rest referring to a subject’s state when instructed to stay awake and not required to perform any given task or pay attention to any specific external stimuli. Those who hypothesize rest dFC to be neuronally relevant have explored the phenomenon in the context of consciousness (Barttfeld et al., 2015), development (Qin et al., 2015) and clinical disorders (Damaraju et al., 2014; Falahpour et al., 2016; Wee et al., 2016). These studies have shown how the complexity of dFC decreases as consciousness levels decline (Barttfeld et al., 2015), how dynamic inter-regional interactions can be used to predict brain maturity (Qin et al., 2015), and how dFC derivatives (e.g., dwell times) can be diagnostically informative for conditions such as schizophrenia (Damaraju et al., 2014), mild cognitive impairment (Wee et al., 2016), and autism (Falahpour et al., 2016). Yet, many others have raised valid concerns regarding the ability of current dFC estimation methods to capture neuronally relevant dFC at rest (Handwerker et al., 2012; Hindriks et al., 2016; Leonardi and Van De Ville, 2015; Shakil et al., 2016). These concerns include a lack of appropriate null models to discern real dynamics from sampling variability (Hindriks et al., 2016), improper pre-processing leading to spurious dynamics (Leonardi and Van De Ville, 2015), and excessive temporal smoothing that hinder our ability to capture sharp and rapid transitions of interest (a real concern for sliding window techniques used to estimate FC states (Keilholz et al., 2017; Shakil et al., 2016);). Finally, some have stated that resting dFC is primarily a manifestation of sampling variability, residual head motion artifacts, and fluctuations in sleep state; and that as such, it lacks clear cognitive or psychological significance (Laumann et al., 2017).

One cause of such discrepant views is that it is challenging to demonstrate the potential cognitive correlates of resting dFC; especially given the unconstrained cognitive nature of rest and scarcity of methods to reliably infer the cognitive correlates of whole-brain FC patterns. When subjects are instructed to quietly rest, retrospective reports demonstrate that subjects often engage in a succession of self-paced cognitive processes including inner speech, musical experience, visual imagery, episodic memory recall, future planning, mental manipulation of numbers, and periods of heightened somatosensory sensations (Delamillieure et al., 2010). Reconfigurations of FC patterns during rest could, to some extent, be a manifestation of this flow of covert self-paced conscious cognition (Barttfeld et al., 2015); even if other factors such as random exploration of cognitive architectures (Deco et al., 2013), fluctuations in autonomic system activity (Chang et al., 2013), intrinsic unconscious processes (Kucyi, 2018), self-initiated body motion (Tan et al., 2017) and arousal levels (Laumann et al., 2017), also contribute. Here we present a series of experiments aimed at exploring this hypothesis, namely that distinct periods of covert cognition are significant contributors to observable resting dFC. We use the term “covert cognition” in the previous sentence as an umbrella term to refer to all different mental states a subject undertakes while lying still inside the MRI scanner. These will differ in nature, order, timing and length across subjects; and may include both ongoing subjective experiences as well as spontaneous memory replays or fluctuations in attention, among many others. In other words, we use the term “cognition during rest” as an equivalent concept to William James “flights” and “perchings” of the mind; and as such, it encompasses the different types of spontaneous thoughts—namely daydreaming, mind-wandering, and creative thinking—described by Christoff et al. (2016) when describing a neuroscientific framework for the study of mind-wandering. In parallel, we also evaluate the ability of FC states—one of the most prominent approaches to summarize rest dFC—to capture those hypothesized periods of distinct cognition during rest.

To explain these diverse and sometimes contradictory observations regarding dFC, we extend current FC state methodology in two ways. First, we combine hemodynamic deconvolution (Caballero Gaudes et al., 2013) and activity-based reverse-inference (Yarkoni et al., 2011) to map FC states onto cognitive states. An FC state is commonly defined in terms of a representative FC matrix and a timeline (when it occurs). The deconvolution step allows us to also generate a representative “activity” map per FC state. Those FC state “activity” maps are subsequently input to a pre-existing decoding framework (Neurosynth (Yarkoni et al., 2011);) able to map whole-brain activity patterns into ranked lists of cognitive processes likely associated with the input map. A direct mapping from representative FC matrices to cognitive processes might be preferable but, unfortunately as of today, such systems do not exist. In the past, hemodynamic deconvolution and reverse-inference have proven successful at decoding the nature of self-driven body motion during rest scans (Tan et al., 2017). Here, we extend that approach for the purpose of attaining open-ended cognitive decoding of FC states. The second way in which we extend current FC state methodology is the use of manifold learning techniques to generate low dimensional representations of time-varying FC that help visualize how connectivity evolves and identify periods of distinct cognition during rest. Projection of dynamic FC patterns into low dimensional spaces to ease interpretation has also been proposed by Dodero et al. (2016), as well as by Shine et al. (2019) to explore system-wide brain dynamics as a function of cognition.

The proposed methods are first evaluated on a multi-task1 fMRI dataset (Gonzalez-Castillo et al., 2015) for which precise information about the timing and nature of cognitive states (as dictated by task demands) is available. Next, we apply the same methods to 15 min long resting state scans with low head motion and infrequent eye closures from the human connectome project (HCP (Van Essen et al., 2013),). This way, we examine the possible presence of periods of distinct cognition during rest, and the ability of FC states to uncover such periods. Comparative analyses across both scenarios (multi-task and rest) are important as substantial differences in the timing and nature of externally driven versus self-paced cognitive processes, and their manifestations in terms of dFC phenomena, may modulate the ability of models to capture cognitively relevant information across both scenarios.

Our results confirm both that FC states can accurately capture periods of distinct cognition driven by external task demands, as previously shown (Gonzalez-Castillo et al., 2015), and also that the proposed extension for the FC state framework permits accurate inference of the cognitive processes underlying each task FC state. Next, we show that although periods of distinct task-like cognition are present during pure rest, and their cognitive correlates can be inferred in a similar manner, FC state modeling—at least in the specific form examined here—does not identify discrete cognitive states as consistently as with the task data. Overall, our work suggests that resting dFC is influenced by periods of task-like cognition, and therefore more than a single correlation structure may be necessary to entirely describe rest scans. At the same time, our results suggest that FC states do not consistently capture periods of distinct cognition during rest, and that their estimation and interpretation as a model for rest dFC should be updated accordingly.

Section snippets

Multi-task dataset

The multi-task dataset used here has been previously described in detail in (Gonzalez-Castillo et al., 2015). In summary, it contains data from 22 subjects (13 females; age 27 ± 5 y.o.) who gave informed consent in compliance with a protocol approved by the Institutional Review Board of the National Institute of Mental Health in Bethesda, MD. The data from two subjects were discarded from the analysis due to excessive spatial distortions in the functional time series.

The MRI data have been

FC states and cognition in multi-task scenario

After pre-processing, FC states were estimated for the multi-task dataset. We performed these analyses under two different scenarios: (a) using only task homogenous windows—namely those that fall completely within the temporal span of individual task blocks (as was previously published (Gonzalez-Castillo et al., 2015)); and (b) using all available windows, which also include windows spanning more than one task block. The goal of comparing these two scenarios is to evaluate to what extent having

Discussion

Resting dFC is an empirically observed phenomenon with promising translational value (Calhoun et al., 2014; Gonzalez-Castillo and Bandettini, 2018) for which several methodological and mechanistic questions remain unresolved (Keilholz et al., 2017). We address some of them here. First, we show evidence in support of the hypothesis that on-going self-paced cognition is a contributor to dFC phenomena during rest, although the degree of contribution may vary across subjects, as well as across

Declaration of interest

The authors declare no competing interests.

Acknowledgements

This research was possible thanks to the support of the National Institute of Mental Health Intramural Research Program. Portions of this study used the high-performance computational capabilities of the Biowulf Linux cluster at the National Institutes of Health, Bethesda, MD (biowulf.nih.gov). This study is part of NIH clinical protocol number NCT00001360, annual report ZIAMH002783, and protocol ID 93-M-0170. CCG was supported by the Spanish Ministry of Economy and Competitiveness through the

References (82)

  • D.A. Handwerker et al.

    Periodic changes in fMRI connectivity

    Neuroimage

    (2012)
  • E.C. Hansen et al.

    Functional connectivity dynamics: modeling the switching behavior of the resting state

    Neuroimage

    (2015)
  • R. Hindriks et al.

    Can sliding-window correlations reveal dynamic functional connectivity in resting-state fMRI?

    Neuroimage

    (2016)
  • R.M. Hutchison et al.

    Dynamic functional connectivity: promise, issues, and interpretations

    Neuroimage

    (2013)
  • H.J. Jo et al.

    Mapping sources of correlation in resting state FMRI, with artifact detection and removal

    Neuroimage

    (2010)
  • F.I. Karahanoglu et al.

    Total activation: fMRI deconvolution through spatio-temporal regularization

    Neuroimage

    (2013)
  • A. Kucyi

    Just a thought: how mind-wandering is represented in dynamic brain connectivity

    Neuroimage

    (2018)
  • N. Leonardi et al.

    Principal components of functional connectivity: a new approach to study dynamic brain connectivity during rest

    Neuroimage

    (2013)
  • N. Leonardi et al.

    On spurious and real fluctuations of dynamic functional connectivity during rest

    Neuroimage

    (2015)
  • M.A. Lindquist et al.

    Evaluating dynamic bivariate correlations in resting-state fMRI: a comparison study and a new approach

    Neuroimage

    (2014)
  • R.A. Poldrack

    Inferring mental states from neuroimaging data: from reverse inference to large-scale decoding

    Neuron

    (2011)
  • M.G. Preti et al.

    The dynamic functional connectome: state-of-the-art and perspectives

    Neuroimage

    (2017)
  • G. Salimi-Khorshidi et al.

    Automatic denoising of functional MRI data: combining independent component analysis and hierarchical fusion of classifiers

    Neuroimage

    (2014)
  • S. Shakil et al.

    Evaluation of sliding window correlation performance for characterizing dynamic functional connectivity and brain states

    Neuroimage

    (2016)
  • A. Tusche et al.

    Classifying the wandering mind: revealing the affective content of thoughts during task-free rest periods

    Neuroimage

    (2014)
  • D.C. Van Essen et al.

    The Wu-minn human connectome project: an overview

    Neuroimage

    (2013)
  • A. Abraham et al.

    Machine learning for neuroimaging with scikit-learn

    Front. Neuroinf.

    (2014)
  • E.A. Allen et al.

    Tracking whole-brain connectivity dynamics in the resting state

    Cerebr. Cortex

    (2014)
  • P. Barttfeld et al.

    Signature of consciousness in the dynamics of resting-state brain activity

    Proc. Natl. Acad. Sci. U. S. A.

    (2015)
  • M. Belkin et al.

    Laplacian Eigenmaps for Dimensionality Reduction and Data Representation

    (2003)
  • V.D. Blondel et al.

    Fast unfolding of communities in large networks

    J. Stat. Mech. Theory Exp.

    (2008)
  • C. Caballero Gaudes et al.

    Paradigm free mapping with sparse regression automatically detects single-trial functional magnetic resonance imaging blood oxygenation level dependent responses

    Hum. Brain Mapp.

    (2013)
  • C. Caballero-Gaudes et al.

    A deconvolution algorithm for multi-echo functional MRI: multi-echo Sparse Paradigm Mappoing

    (2019)
  • E. Chah et al.

    Automated spike sorting algorithm based on Laplacian eigenmaps and k-means clustering

    J. Neural Eng.

    (2011)
  • J.M. Chambers

    Graphical Methods for Data Analysis

    (1983)
  • C. Chang et al.

    Tracking brain arousal fluctuations with fMRI

    Proc. Natl. Acad. Sci. U. S. A.

    (2016)
  • H. Chernoff

    The use of faces to represent points in k-dimensional space graphically

    J. Am. Stat. Assoc.

    (1973)
  • K. Christoff et al.

    Mind-wandering as spontaneous thought: a dynamic framework

    Nat. Rev. Neurosci.

    (2016)
  • R.C. Craddock et al.

    A whole brain fMRI atlas generated via spatially constrained spectral clustering

    Hum. Brain Mapp.

    (2012)
  • X. Di et al.

    Task vs. rest-different network configurations between the coactivation and the resting-state brain networks

    Front. Hum. Neurosci.

    (2013)
  • B.A. Diaz et al.

    The Amsterdam Resting-State Questionnaire reveals multiple phenotypes of resting-state cognition

    Front. Hum. Neurosci.

    (2013)
  • Cited by (37)

    • Neural representations of self-generated thought during think-aloud fMRI

      2023, NeuroImage
      Citation Excerpt :

      Buckner et al. noted that while resting-state functional connectivity (FC) is not entirely constrained by anatomic connectivity, it is influenced by the mental state of the participant during scanning (Buckner et al., 2013). By comparing dynamic FC analyses between resting and anesthetized states, Barttfeld et al. found that patterns of FC during the resting state also reflected ongoing cognitive processing (Barttfeld et al., 2015); Gonzalez-Castillo et al. found that resting-state dynamic FC was influenced by short periods of spontaneous cognitive-task-like processes (Gonzalez-Castillo et al., 2019). Thus, ongoing experience (self-generated thought) during resting-state fMRI affects the detected signals.

    View all citing articles on Scopus
    View full text