The functional connectivity landscape of the human brain

Abstract

Functional brain networks emerge and dissipate over a primarily static anatomical foundation. The dynamic basis of these networks is inter-regional communication involving local and distal regions. It is assumed that inter-regional distances play a pivotal role in modulating network dynamics. Using three different neuroimaging modalities, 6 datasets were evaluated to determine whether experimental manipulations asymmetrically affect functional relationships based on the distance between brain regions in human participants. Contrary to previous assumptions, here we show that short- and long-range connections are equally likely to strengthen or weaken in response to task demands. Additionally, connections between homotopic areas are the most stable and less likely to change compared to any other type of connection. Our results point to a functional connectivity landscape characterized by fluid transitions between local specialization and global integration. This ability to mediate functional properties irrespective of spatial distance may engender a diverse repertoire of cognitive processes when faced with a dynamic environment.

Figure 1. Fisher r-to-z-transformed correlation matrices for all studies exhibiting the correlation between all nodes used to analyze each study.

For the PLS analyses, each participant’s matrix was used in the analysis.

Figure 2. Relationship between the Euclidean distance of parcels and the correlation between parcel timeseries for each fMRI study and for each condition and participant group.

Figure 3. First Latent Variables (LVs) from each of the four fMRI Datasets as uncovered with PLS.

The significance of the LVs and the amount of cross-block covariance explained are listed in the titles.

Figure 4. PLS Saliences for the first latent variable in each study plotted against Euclidean distance of each functional connection.

Figure 5. Changes in correlation between conditions of greatest difference in correlation plotted against Euclidean distance.

Figure 6. PLS bootstrap ratios for each functional connection are plotted against the Euclidean distance of that connection.

Figure 7. 3-dimensional histogram of correlation differences by Euclidean distance for each study.

In the 3-dimensional histogram (top) the number of observations (i.e., density) is represented by color and height. In the 2-dimensional plot (bottom) density is represented by color only.

Figure 8. PCA analysis of scatter plots displaying the relationship between correlation differences and Euclidean distance.

The PCA analysis was used to determine if outlying points exhibited a different relationship with Euclidean distance than the majority of the points. Points that were in the upper or lower 5% in terms of displacement from the primary axis of variance are colored red and suggest that there is no systematic relationship between correlation differences and Euclidean distances for the outlying points. The red line represents the primary axis of variance.

Figure 9. Absolute value of the Saliences for each of the four fMRI studies plotted against Euclidean distance.

Figure 10. Absolute value of the correlation differences for each of the four fMRI studies plotted against Euclidean distance.

Figure 11. Relationship between Euclidean distance and Salience for functional connections within and between modules.

Figure 12. Relationship between changes in coherence and Euclidean distance for the MEG dataset (left) and differences in path coefficients from an SEM analysis of the PET dataset (right).

Figure 13. Relationship between changes in functional connectivity for nodes within a hemisphere (A), between hemispheres (B), between homotopic areas (C) and between non-topic areas (D).

(E) Histogram showing the changes in functional connectivity for homotopic and non-homotopic areas.