Instantaneous and causal connectivity in resting state brain networks derived from functional MRI data

Abstract

Background: Most neuroimaging studies of resting state networks have concentrated on functional connectivity (FC) based on instantaneous correlation in a single network. In this study we investigated both FC and effective connectivity (EC) based on Granger causality of four important networks at resting state derived from functional magnetic resonance imaging data - default mode network (DMN), hippocampal cortical memory network (HCMN), dorsal attention network (DAN) and fronto-parietal control network (FPCN).

Methodology/principle findings: A method called correlation-purged Granger causality analysis was used, not only enabling the simultaneous evaluation of FC and EC of all networks using a single multivariate model, but also accounting for the interaction between them resulting from the smoothing of neuronal activity by hemodynamics. FC was visualized using a force-directed layout upon which causal interactions were overlaid. FC results revealed that DAN is very tightly coupled compared to the other networks while the DMN forms the backbone around which the other networks amalgamate. The pattern of bidirectional causal interactions indicates that posterior cingulate and posterior inferior parietal lobule of DMN act as major hubs. The pattern of unidirectional causal paths revealed that hippocampus and anterior prefrontal cortex (aPFC) receive major inputs, likely reflecting memory encoding/retrieval and cognitive integration, respectively. Major outputs emanating from anterior insula and middle temporal area, which are directed at aPFC, may carry information about interoceptive awareness and external environment, respectively, into aPFC for integration, supporting the hypothesis that aPFC-seeded FPCN acts as a control network.

Conclusions/significance: Our findings indicate the following. First, regions whose activities are not synchronized interact via time-delayed causal influences. Second, the causal interactions are organized such that cingulo-parietal regions act as hubs. Finally, segregation of different resting state networks is not clear cut but only by soft boundaries.

Figure 1

A force-directed layout of the zero-lag correlation network obtained from 33 ROIs in resting state using spring embedding algorithm. The significant (p<0.01) instantaneous correlation between the ROIs is shown as green lines. The distance between the nodes indicates the strength of their connection. The correspondence between ROIfontcolors and the RSNs is as follows. Red: Default mode network (DMN), Green: Dorsal attention network (DAN), Blue: Hippocampal-cortical memory network (HCMN) and Black: Fronto-parietal control network (FPCN). The correspondence between ROIbackgroundcolors and their anatomical location is as follows. Yellow: Frontal cortex, Pink: Parietal cortex, Blue: Temporal cortex and Green: Cingulate cortex.

Figure 2

The significant (p<0.01) bi-directional causal paths of the CPGC network overlaid on the force-directed layout of instantaneous correlation network. The reciprocal causal connections are shown as blue lines and the instantaneous correlation as green lines. The causal network formed by the blue lines represents the only strongly connected component in the entire network where in every ROI is mutually accessible causally by every other ROI. The correspondence between ROIfontcolors and the RSNs is as follows. Red: Default mode network (DMN), Green: Dorsal attention network (DAN), Blue: Hippocampal-cortical memory network (HCMN) and Black: Fronto-parietal control network (FPCN). The correspondence between ROIbackgroundcolors and their anatomical location is as follows. Yellow: Frontal cortex, Pink: Parietal cortex, Blue: Temporal cortex and Green: Cingulate cortex.

Figure 3

The significant (p<0.01) paths with unidirectional influences in the CPGC network. The width of the blue lines is proportional to the strength of the corresponding path. The correspondence between ROIfontcolors and the RSNs is as follows. Red: Default mode network (DMN), Green: Dorsal attention network (DAN), Blue: Hippocampal-cortical memory network (HCMN) and Black: Fronto-parietal control network (FPCN). The correspondence between ROIbackgroundcolors and their anatomical location is as follows. Yellow: Frontal cortex, Pink: Parietal cortex, Blue: Temporal cortex and Green: Cingulate cortex.