Dynamic causal modelling of fluctuating connectivity in resting-state EEG

Frederik Van de Steen, Hannes Almgren, Adeel Razi, Karl Friston, Daniele Marinazzo

Research output: Contribution to journalArticleResearchpeer-review

1 Citation (Scopus)

Abstract

Functional and effective connectivity are known to change systematically over time. These changes might be explained by several factors, including intrinsic fluctuations in activity-dependent neuronal coupling and contextual factors, like experimental condition and time. Furthermore, contextual effects may be subject-specific or conserved over subjects. To characterize fluctuations in effective connectivity, we used dynamic causal modelling (DCM) of cross spectral responses over 1- min of electroencephalogram (EEG) recordings during rest, divided into 1-sec windows. We focused on two intrinsic networks: the default mode and the saliency network. DCM was applied to estimate connectivity in each time-window for both networks. Fluctuations in DCM connectivity parameters were assessed using hierarchical parametric empirical Bayes (PEB). Within-subject, between-window effects were modelled with a second-level linear model with temporal basis functions as regressors. This procedure was conducted for every subject separately. Bayesian model reduction was then used to assess which (combination of) temporal basis functions best explain dynamic connectivity over windows. A third (between-subject) level model was used to infer which dynamic connectivity parameters are conserved over subjects. Our results indicate that connectivity fluctuations in the default mode network and to a lesser extent the saliency network comprised both subject-specific components and a common component. For both networks, connections to higher order regions appear to monotonically increase during the 1- min period. These results not only establish the predictive validity of dynamic connectivity estimates – in virtue of detecting systematic changes over subjects – they also suggest a network-specific dissociation in the relative contribution of fluctuations in connectivity that depend upon experimental context. We envisage these procedures could be useful for characterizing brain state transitions that may be explained by their cognitive or neuropathological underpinnings.

Original languageEnglish
Pages (from-to)476-484
Number of pages9
JournalNeuroImage
Volume189
DOIs
Publication statusPublished - 1 Apr 2019

Keywords

  • Dynamic causal modelling
  • EEG
  • Fluctuating connectivity
  • Resting state

Cite this

Van de Steen, Frederik ; Almgren, Hannes ; Razi, Adeel ; Friston, Karl ; Marinazzo, Daniele. / Dynamic causal modelling of fluctuating connectivity in resting-state EEG. In: NeuroImage. 2019 ; Vol. 189. pp. 476-484.
@article{b84f13c32de44f878f9c17b7d085df67,
title = "Dynamic causal modelling of fluctuating connectivity in resting-state EEG",
abstract = "Functional and effective connectivity are known to change systematically over time. These changes might be explained by several factors, including intrinsic fluctuations in activity-dependent neuronal coupling and contextual factors, like experimental condition and time. Furthermore, contextual effects may be subject-specific or conserved over subjects. To characterize fluctuations in effective connectivity, we used dynamic causal modelling (DCM) of cross spectral responses over 1- min of electroencephalogram (EEG) recordings during rest, divided into 1-sec windows. We focused on two intrinsic networks: the default mode and the saliency network. DCM was applied to estimate connectivity in each time-window for both networks. Fluctuations in DCM connectivity parameters were assessed using hierarchical parametric empirical Bayes (PEB). Within-subject, between-window effects were modelled with a second-level linear model with temporal basis functions as regressors. This procedure was conducted for every subject separately. Bayesian model reduction was then used to assess which (combination of) temporal basis functions best explain dynamic connectivity over windows. A third (between-subject) level model was used to infer which dynamic connectivity parameters are conserved over subjects. Our results indicate that connectivity fluctuations in the default mode network and to a lesser extent the saliency network comprised both subject-specific components and a common component. For both networks, connections to higher order regions appear to monotonically increase during the 1- min period. These results not only establish the predictive validity of dynamic connectivity estimates – in virtue of detecting systematic changes over subjects – they also suggest a network-specific dissociation in the relative contribution of fluctuations in connectivity that depend upon experimental context. We envisage these procedures could be useful for characterizing brain state transitions that may be explained by their cognitive or neuropathological underpinnings.",
keywords = "Dynamic causal modelling, EEG, Fluctuating connectivity, Resting state",
author = "{Van de Steen}, Frederik and Hannes Almgren and Adeel Razi and Karl Friston and Daniele Marinazzo",
year = "2019",
month = "4",
day = "1",
doi = "10.1016/j.neuroimage.2019.01.055",
language = "English",
volume = "189",
pages = "476--484",
journal = "NeuroImage",
issn = "1053-8119",
publisher = "Elsevier",

}

Dynamic causal modelling of fluctuating connectivity in resting-state EEG. / Van de Steen, Frederik; Almgren, Hannes; Razi, Adeel; Friston, Karl; Marinazzo, Daniele.

In: NeuroImage, Vol. 189, 01.04.2019, p. 476-484.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Dynamic causal modelling of fluctuating connectivity in resting-state EEG

AU - Van de Steen, Frederik

AU - Almgren, Hannes

AU - Razi, Adeel

AU - Friston, Karl

AU - Marinazzo, Daniele

PY - 2019/4/1

Y1 - 2019/4/1

N2 - Functional and effective connectivity are known to change systematically over time. These changes might be explained by several factors, including intrinsic fluctuations in activity-dependent neuronal coupling and contextual factors, like experimental condition and time. Furthermore, contextual effects may be subject-specific or conserved over subjects. To characterize fluctuations in effective connectivity, we used dynamic causal modelling (DCM) of cross spectral responses over 1- min of electroencephalogram (EEG) recordings during rest, divided into 1-sec windows. We focused on two intrinsic networks: the default mode and the saliency network. DCM was applied to estimate connectivity in each time-window for both networks. Fluctuations in DCM connectivity parameters were assessed using hierarchical parametric empirical Bayes (PEB). Within-subject, between-window effects were modelled with a second-level linear model with temporal basis functions as regressors. This procedure was conducted for every subject separately. Bayesian model reduction was then used to assess which (combination of) temporal basis functions best explain dynamic connectivity over windows. A third (between-subject) level model was used to infer which dynamic connectivity parameters are conserved over subjects. Our results indicate that connectivity fluctuations in the default mode network and to a lesser extent the saliency network comprised both subject-specific components and a common component. For both networks, connections to higher order regions appear to monotonically increase during the 1- min period. These results not only establish the predictive validity of dynamic connectivity estimates – in virtue of detecting systematic changes over subjects – they also suggest a network-specific dissociation in the relative contribution of fluctuations in connectivity that depend upon experimental context. We envisage these procedures could be useful for characterizing brain state transitions that may be explained by their cognitive or neuropathological underpinnings.

AB - Functional and effective connectivity are known to change systematically over time. These changes might be explained by several factors, including intrinsic fluctuations in activity-dependent neuronal coupling and contextual factors, like experimental condition and time. Furthermore, contextual effects may be subject-specific or conserved over subjects. To characterize fluctuations in effective connectivity, we used dynamic causal modelling (DCM) of cross spectral responses over 1- min of electroencephalogram (EEG) recordings during rest, divided into 1-sec windows. We focused on two intrinsic networks: the default mode and the saliency network. DCM was applied to estimate connectivity in each time-window for both networks. Fluctuations in DCM connectivity parameters were assessed using hierarchical parametric empirical Bayes (PEB). Within-subject, between-window effects were modelled with a second-level linear model with temporal basis functions as regressors. This procedure was conducted for every subject separately. Bayesian model reduction was then used to assess which (combination of) temporal basis functions best explain dynamic connectivity over windows. A third (between-subject) level model was used to infer which dynamic connectivity parameters are conserved over subjects. Our results indicate that connectivity fluctuations in the default mode network and to a lesser extent the saliency network comprised both subject-specific components and a common component. For both networks, connections to higher order regions appear to monotonically increase during the 1- min period. These results not only establish the predictive validity of dynamic connectivity estimates – in virtue of detecting systematic changes over subjects – they also suggest a network-specific dissociation in the relative contribution of fluctuations in connectivity that depend upon experimental context. We envisage these procedures could be useful for characterizing brain state transitions that may be explained by their cognitive or neuropathological underpinnings.

KW - Dynamic causal modelling

KW - EEG

KW - Fluctuating connectivity

KW - Resting state

UR - http://www.scopus.com/inward/record.url?scp=85060860423&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2019.01.055

DO - 10.1016/j.neuroimage.2019.01.055

M3 - Article

VL - 189

SP - 476

EP - 484

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

ER -