Higher Frequency Network Activity Flow Predicts Lower Frequency Node Activity in Intrinsic Low-Frequency BOLD Fluctuations

The brain remains electrically and metabolically active during resting conditions. The low-frequency oscillations (LFO) of the blood oxygen level-dependent (BOLD) signal of functional magnetic resonance imaging (fMRI) coherent across distributed brain regions are known to exhibit features of this activity. However, these intrinsic oscillations may undergo dynamic changes in time scales of seconds to minutes during resting conditions. Here, using wavelet-transform based time-frequency analysis techniques, we investigated the dynamic nature of default-mode networks from intrinsic BOLD signals recorded from participants maintaining visual fixation during resting conditions. We focused on the default-mode network consisting of the posterior cingulate cortex (PCC), the medial prefrontal cortex (mPFC), left middle temporal cortex (LMTC) and left angular gyrus (LAG). The analysis of the spectral power and causal flow patterns revealed that the intrinsic LFO undergo significant dynamic changes over time. Dividing the frequency interval 0 to 0.25 Hz of LFO into four intervals slow-5 (0.01-0.027 Hz), slow-4 (0.027-0.073 Hz), slow-3 (0.073-0.198 Hz) and slow-2 (0.198-0.25 Hz), we further observed significant positive linear relationships of slow-4 in-out flow of network activity with slow-5 node activity, and slow-3 in-out flow of network activity with slow-4 node activity. The network activity associated with respiratory related frequency (slow-2) was found to have no relationship with the node activity in any of the frequency intervals. We found that the net causal flow towards a node in slow-3 band was correlated with the number of fibers, obtained from diffusion tensor imaging (DTI) data, from the other nodes connecting to that node. These findings imply that so-called resting state is not 'entirely' at rest, the higher frequency network activity flow can predict the lower frequency node activity, and the network activity flow can reflect underlying structural connectivity.