Investigating delayed functional connectivity networks in the human brain

MRI-based functional connectivity is a widely used method for studying the human brain that is based upon the cross-correlations between time series from different brain regions. Traditionally, this method has been based upon cross-correlations that have no time difference between the time series (i.e., zero-lag correlation). More recently, a number of factors have suggested that time delayed correlations between brain regions may be functionally meaningful. These include a growing understanding of the BOLD signal, identification of temporal structure within brain-wide activity, and the identification of intrinsic timescales in different brain regions. When trying to study these temporal features through BOLD fMRI, it is often necessary to up-sample data in order to estimate effects at a better resolution than is provided by the raw data acquisition. In theory, this upsampling of functional connectivity could be done before the correlation step in the functional connectivity calculation is done, or after. Both approaches have been used in different contexts. The purpose of this study was to see if this methodological choice influences the results obtained. To do this, we use two different resting-state datasets, one of which includes two runs from the same participants. With these datasets we investigate the reliability of lag thread and seed based functional connectivity patterns and temporal delays within subjects and across datasets. We find that up-sampling after the correlation step produces more robust results for all measures. These results provide useful methodological information for those seeking to study functional connectivity delays with BOLD fMRI.