Dynamic fluctuations in global brain network topology characterize functional states during rest and behavior

James M. Shine, Peter T. Bell, Oluwasanmi Koyejo, Patrick G. Bissett, Krzysztof J. Gorgolewski, Craig A. Moodie, Russell A. Poldrack

Abstract

Higher brain function relies upon the ability to flexibly integrate information across specialized communities of macroscopic brain regions, but it is unclear how this mechanism manifests over time. Here we characterized patterns of time-resolved functional connectivity using resting state and task fMRI data from a large cohort of unrelated subjects. Our results demonstrate that dynamic fluctuations in network structure during the resting state reflect transitions between states of integrated and segregated network topology. These patterns were altered during task performance, demonstrating a higher level of network integration that tracked with the complexity of the task and correlated with effective behavioral performance. Replication analysis demonstrated that these results were reproducible across sessions, sample populations and datasets. Together these results provide insight into the brain's coordination between integration and segregation and highlight key principles underlying the reorganization of the network architecture of the brain during both rest and behavior.

Introduction

For decades, neuroscience viewed brain networks as relatively static maps. While specialized regions for vision, language, or movement were well-defined, the mechanism by which these "neighborhoods" coordinate in real-time to produce complex thought remained elusive. A central challenge in brain architecture is the tension between segregation—efficiency through specialization—and integration—flexibility through global collaboration. This study investigates whether the brain's global architecture dynamically shifts between these modes and whether such transitions underpin psychological performance.

Methods

To capture the brain in motion, the authors utilized high-resolution functional connectivity estimation and network topology metrics across large-scale datasets, including the Human Connectome Project (HCP) and the NKI Rockland Sample.

Analytical Framework

  • Multiplication of Temporal Derivatives (MTD): Used to estimate functional connectivity at high temporal resolution, avoiding the blurring effects of standard sliding-window correlations.
  • Cartographic Profiling: Every brain region was mapped based on its within-module degree ($W_T$) and its participation coefficient ($B_T$), which measures its connectivity across different communities.
  • Unsupervised Clustering: k-means clustering was applied to these profiles to identify naturally occurring brain states.
  • Pupillometry: Pupil diameter served as a proxy for the activity of the Locus Coeruleus, linking network topology to the brain's internal arousal systems.
Figure 1: Methodological pipeline and state identification. (a) Time series of mean participation coefficient ($B_T$) showing fluctuations between states. (b) Density plots of $W_T$ and $B_T$ defining the Segregated and Integrated states. (c-d) Spatial maps of changes in $W_T$ and $B_T$ during state transitions.

Results

The analysis revealed that the brain is metastable, aperiodically flipping between two primary topological states.

The Integrated vs. Segregated States

The resting brain spends approximately 70% of its time in an integrated state, characterized by high global connectivity. In contrast, the segregated state is defined by localized, specialized processing. These states are statistically distinct in their topological profiles.

Task-Induced Reorganization

When subjects performed a challenging "N-back" memory task, the brain underwent a global shift toward the integrated state. This shift was most pronounced in "hub" regions—a right-lateralized network of frontoparietal and subcortical areas—which facilitate communication between specialist regions.

Figure 2: Spatial distribution of topological changes. (c) Changes in within-module degree ($\Delta W_T$) and (d) changes in participation coefficient ($\Delta B_T$) highlighting the regions that facilitate global integration.
  • Performance: Higher integration during the task correlated significantly with faster information processing (drift rates, $v$) and shorter non-decision times ($t$).
  • Arousal: Pupil diameter showed a positive correlation ($r = 0.241$) with between-module connectivity ($B_T$), suggesting that neuromodulatory systems like norepinephrine act as a "volume knob" for global integration.

Discussion

These findings suggest that the brain's ability to integrate information is a dynamic process regulated by internal arousal states. While the use of pupillometry provides a compelling link to neuromodulation, direct electrophysiological confirmation is needed. Future research should explore whether specific cognitive sub-components rely on distinct topological signatures and how individual differences in functional boundaries affect these global network shifts.

Summary updated

Abstract

Higher brain function relies upon the ability to flexibly integrate information across specialized communities of macroscopic brain regions, but it is unclear how this mechanism manifests over time. Here we characterized patterns of time-resolved functional connectivity using resting state and task fMRI data from a large cohort of unrelated subjects. Our results demonstrate that dynamic fluctuations in network structure during the resting state reflect transitions between states of integrated and segregated network topology. These patterns were altered during task performance, demonstrating a higher level of network integration that tracked with the complexity of the task and correlated with effective behavioral performance. Replication analysis demonstrated that these results were reproducible across sessions, sample populations and datasets. Together these results provide insight into the brain's coordination between integration and segregation and highlight key principles underlying the reorganization of the network architecture of the brain during both rest and behavior.
BibTeX
@inproceedings{shine2016dynamic,
  author = {James M. Shine and Peter T. Bell and Oluwasanmi Koyejo and Patrick G. Bissett and Krzysztof J. Gorgolewski and Craig A. Moodie and Russell A. Poldrack},
  title = {{Dynamic fluctuations in global brain network topology characterize functional states during rest and behavior}},
  booktitle = {Neuron},
  year = {2016},
  doi = {10.1016/j.neuron.2016.09.018},
}
Figure