Monday, October 7, 2019

Individual variations in the modular organization of functional brain networks: higher intelligence seems associated with higher temporal stability (lower temporal variability) of brain network modularity

Temporal stability of functional brain modules associated with human intelligence. Kirsten Hilger et al. Human Brain Mapping, October 6 2019.

Abstract: Individual differences in general cognitive ability (i.e., intelligence) have been linked to individual variations in the modular organization of functional brain networks. However, these analyses have been limited to static (time‐averaged) connectivity, and have not yet addressed whether dynamic changes in the configuration of brain networks relate to general intelligence. Here, we used multiband functional MRI resting‐state data (N = 281) and estimated subject‐specific time‐varying functional connectivity networks. Modularity optimization was applied to determine individual time‐variant module partitions and to assess fluctuations in modularity across time. We show that higher intelligence, indexed by an established composite measure, the Wechsler Abbreviated Scale of Intelligence (WASI), is associated with higher temporal stability (lower temporal variability) of brain network modularity. Post‐hoc analyses reveal that subjects with higher intelligence scores engage in fewer periods of extremely high modularity — which are characterized by greater disconnection of task‐positive from task‐negative networks. Further, we show that brain regions of the dorsal attention network contribute most to the observed effect. In sum, our study suggests that investigating the temporal dynamics of functional brain network topology contributes to our understanding of the neural bases of general cognitive abilities.


Intelligence describes our ability to reason, to understand complex ideas, to learn from experiences, and to adapt effectively to the environment (Neisser et al., 1996). Understanding the biological bases of human intelligence is an important scientific aim, and neuroscientific research has begun to contribute insights about how individual differences in brain function (Duncan, 2005; Sripada, Angstadt, & Rutherford, 2018), brain structure (Gregory et al., 2016; Haier, Jung, Yeo, Head, & Alkire, 2004), and intrinsic brain connectivity (Hilger, Ekman, Fiebach, & Basten, 2017a; Van den Heuvel, Stam, Kahn, & Hulshoff Pol, 2009) relate to general intelligence (for review see Basten, Hilger, & Fiebach, 2015; Jung & Haier, 2007).

Recent years have seen an increasing interest in understanding how human cognition emerges from the intrinsic organization of functional brain networks (Park & Friston, 2013), often studied using functional MRI (fMRI) in the absence of task demands (i.e., under resting‐state conditions; Biswal, Yetkin, Haughton, & Hyde, 1995). The topology of these networks determines how information is transferred between brain regions, and graph theory provides a set of tools to study these topological characteristics (Rubinov & Sporns, 2010). In the field of intelligence research, early graph‐theoretical work proposed that global properties of brain networks such as higher global network efficiency are associated with higher intelligence (van den Heuvel et al., 2009), a finding not replicated in more recent studies (Kruschwitz, Waller, Daedelow, Walter, & Veer, 2018; Pamplona, Santos Neto, Rosset, Rogers, & Salmon, 2015). In contrast, other studies have suggested that intelligence is related to efficiency in the interconnections of specific brain regions (Hilger et al., 2017a). Graph‐theoretical investigations revealed further that the human brain exhibits a hierarchically modular organization with clusters of nodes (modules, subnetworks) that are densely connected among each other but only sparsely coupled to nodes in other modules (Meunier, Lambiotte, & Bullmore, 2010; Sporns & Betzel, 2016). A modular organization balances segregated and integrated information processing, both of which are important for human cognition (Cohen & D'Esposito, 2016). Region‐specific modularity was recently also shown to covary significantly with individual differences in general intelligence (Hilger, Ekman, Fiebach, & Basten, 2017b).

The functional brain network correlates of intelligence were so far mostly studied as a static (i.e., time‐invariant) property of the human brain, that is, by averaging time courses of neural activation across the entire duration of a resting‐state fMRI scan (typically 5–10 min). This approach, however, ignores that intrinsic brain networks vary substantially across time (Cohen, 2018; Lurie et al., 2018; Zalesky, Fornito, Cocchi, Gollo, & Breakspear, 2014). Importantly, it has been shown that the dynamic interplay between states of high integration (low modularity) versus high segregation (high modularity) is linked to different levels of attention (Shine, Koyejo, & Poldrack, 2016) and cognitive performance (Shine et al., 2016). These first results suggest that the study of network dynamics has great potential for providing insights into human cognition from a mechanistic point of view — and thus also for advancing our understanding about the neural mechanisms underlying different levels of general cognitive ability.

Here, we apply graph‐theoretical modularity analyses to resting‐state BOLD fMRI data from a large sample of healthy adult humans (N = 281) to test the hypothesis that intelligence covaries significantly with the amount of dynamic reconfiguration within modularly organized, intrinsic brain networks. Going beyond previous work, we measured global modularity at different spatial scales, to gain insights into the brain's intrinsic network architecture beyond an arbitrarily chosen resolution level. The results of this analysis replicate and extend our previous finding that intelligence is not related to global modularity of static (i.e., time‐invariant) networks (Hilger et al., 2017b). Most importantly, we observed an association between intelligence and dynamic network reconfiguration, such that more intelligent persons show greater stability of network segregation over time.

Data Availability Statement: All data used in the current study can be accessed online under: The preprocessing pipeline CCS is also freely available to the public via GitHub ( or All further analysis code used in the current study has been deposited on GitHub (‐Brain‐Network‐Modularity) and Zenodo (

No comments:

Post a Comment