Kay Jann^{1} and Danny JJ Wang^{1}

While dynamic functional connectivity (dynFC) provides an estimate of the information transfer between brain network nodes, the signal dynamics at each node represents the local information processing. Here we assessed the relation between dynFC within the default mode network and the complexity/regularity of fMRI signal of network nodes. We found that a more complex and thus less predictable signal in a node allows for a more dynamic connectivity and hence a richer repertoire of different FC states.

We
analyzed 31 subjects from the HCP database ^{2}. Data were acquired at 3T with TR/TE = 720/33ms,
multiband-factor 8, FA = 52°, gradient-echo EPI readout and 2mm isotropic
resolution ^{3}. Datasets were preprocessed according to HCP minimal
preprocessing pipeline ^{4}.

Six nodes of the DMN
were defined from a template-DMN ^{5}: ACC/mPFC, PCC, left & right IPL and
left & right hippocampus. First, we computed static FC between all nodes
using conventional Pearson correlation. Second, using a sliding window approach
(window-length 20TRs) the dynamic changes in interregional FC was computed. MSE
was calculated using sample-entropy for 40 coarse-sampled temporal scales
(pattern length m=2, matching threshold r=0.3) ^{6}. This allowed comparing FC
and dynFC to signal complexity at different temporal frequencies. To test for
potential relations between the local signal dynamics (MSE) and network
coherence (FC and dynFC) we (i) correlated the overall network FC with overall
network MSE across scales to identify a global association between network
connectivity and network complexity; (ii) correlated MSE to the average static FC
of each node as well as the variance of dynFC at each node, respectively to
identify the frequency range where MSE correlates to nodal connectivity; and (iii)
calculated the correlation between meanMSE of each node and node-to-node variance
of dynFC to evaluate if there is uni- or bidirectional relation between the
dynamic changes in FC between two nodes and their respective MSE.

Correlation between average static FC of the DMN with
average MSE of all DMN nodes for all frequency scales (red asterisks indicate
statistical significance at p<0.05). At higher scales MSE and FC within the
DMN show a negative correlation suggesting that the more complex the
low-frequency network fluctuations the less connected is the network.

Relation between a node’s meanMSE (across all freuquency
scales) and its dynamic connectivity to other network nodes. Left side shows
the correlation matrix (only significant correlations (p<0.05), diagonal
contains no value). Right plot illustrates the matrix with arrows pointing from
the node whose MSE correlates to the dynFC with the node at the arrowhead. It
can be observed that MSE is related to the variance of dynFC within the areas
of the DMN. Especially the more central a node (hub vs. periphery) is within
the DMN, the stronger the influence of MSE on its dynFC (upper triangle in correlation matrix)