Linking neuronal and hemodynamic network signatures in the resting human brain

Despite several studies investigating the relationship between blood-oxygen-level-dependent functional MRI (BOLD-fMRI) and neuroelectric activity, our understanding is rather incomplete. For instance, the canonical hemodynamic response function (HRF) is commonly used, regardless of brain region, frequency of electric activity and functional networks. We studied this relationship between BOLD-fMRI and electroencephalography (EEG) signal of the human brain in detail using simultaneous fMRI and EEG in healthy awake human subjects at rest. Signals from EEG sensors were filtered into different frequency bands and reconstructed it in the three-dimensional source space. The correlation of the time courses of the two modalities were quantified on a voxel-by-voxel basis on full-brain level as well as separately for each resting state network, with different temporal shifts and EEG frequency bands. We found highly significant correlations between the BOLD-fMRI signal and simultaneously measured EEG, yet with varying time-lags for different frequency bands and different resting state networks. Additionally, we found significant negative correlations with a much longer delay in the fMRI BOLD signal. The positive correlations were mostly around 6-8 seconds delayed in the BOLD time course while the negative correlations were noticed with a BOLD delay of around 20 to 26 seconds. These positive and negative correlation patterns included the commonly reported alpha and gamma bands but also extend in other frequency bands giving characteristic profiles for different resting state networks. Our results confirm recent works that suggest that the relationship between the two modalities is rather brain region / network-specific than a global function and suggest that applying a global canonical HRF for electrophysiological data is probably insufficient to account for the different spatial and temporal dynamics of different brain networks. Moreover, our results suggest that the HRF also varies in different frequency bands giving way to further studies investigating cross-frequency coupling and its interplay with resting state networks.