Establishing the cognitive signature of human brain networks derived from structural and functional connectivity

Abstract

Numerous neuroimaging studies have identified various brain networks using task-free analyses. While these networks undoubtedly support higher cognition, their precise functional characteristics are rarely probed directly. The frontal, temporal, and parietal lobes contain the majority of the tertiary association cortex, which are key substrates for higher cognition including executive function, language, memory, and attention. Accordingly, we established the cognitive signature of a set of contrastive brain networks on the main tertiary association cortices, identified in two task-independent datasets. Using graph-theory analysis, we revealed multiple networks across the frontal, temporal, and parietal cortex, derived from structural and functional connectivity. The patterns of network activity were then investigated using three task-active fMRI datasets to generate the functional profiles of the identified networks. We employed representational dissimilarity analysis on these functional data to quantify and compare the representational characteristics of the networks. Our results demonstrated that the topology of the task-independent networks was strongly associated with the patterns of network activity in the task-active fMRI. Our findings establish a direct relationship between the brain networks identified from task-free datasets and higher cognitive functions including cognitive control, language, memory, visuospatial function, and perception. Not only does this study support the widely held view that higher cognitive functions are supported by widespread, distributed cortical networks, but also it elucidates a methodological approach for formally establishing their relationship.Electronic supplementary materialThe online version of this article (10.1007/s00429-018-1734-x) contains supplementary material, which is available to authorized users.

A Comparison of Functional Networks Derived From Representational Similarity, Functional Connectivity, and Univariate Analyses

“…The high similarity in the neural networks of RSA and FCA provides support for the idea that brain areas showing similar stimulus-related selectivity are also intrinsically connected. Our finding is in line with previous resting-state fMRI studies that have identified functionally relevant networks, such as the primary visual network, auditory network, motor network, and cognitive networks, during rest (e.g., Biswal et al, 1995;Fox and Raichle, 2007;Jung et al, 2018).…”

“…Such distinctions between networks derived from different methods have also been observed in the recent study of Jung and her colleagues (Jung et al, 2018) comparing resting-state fMRI and structural connectivity. Although their comparison involves different methods than ours, they provided some possible explanations that should be considered in the current study.…”

Comparing the functional structure of neural networks from representational similarity analysis with those from functional connectivity and univariate analyses

“…The high similarity in the 491 neural networks of RSA and FCA provides support for the idea that brain areas showing similar 492 stimulus-related selectivity are also intrinsically connected. Our finding is in line with previous 493 resting-state fMRI studies that have identified functionally relevant networks, such as the 494 primary visual network, auditory network, motor network, and cognitive networks, during rest 495 (e.g., Biswal et al, 1995;Fox and Raichle, 2007;Jung et al, 2018). 496…”

“…Such distinctions between the network structures derived from different methods has 523 also been observed in the recent study of Jung and her colleagues (Jung et al, 2018) comparing 524 resting-state fMRI and structural connectivity. Although their comparison involves different 525 methods than ours, they provided some possible explanations that should be considered in the 526 current study.…”

“…Although their comparison involves different 525 methods than ours, they provided some possible explanations that should be considered in the 526 current study. Quality and nature of the datasets used for three methods (even from identical 527 data sources, but measured at different times or analyzed in a different way) may not be equal 528 and different measurement noise may be present (Jung et al, 2018). In addition, they mention 529 that networks during mental activity are modulated away (slightly) from intrinsic connections, 530 which is especially relevant to the comparison of RSA with FCA.…”