Functional brain network organisation of children between 2 and 5 years derived from reconstructed activity of cortical sources of high-density EEG recordings.

There is increasing interest in applying connectivity analysis to brain measures (Rubinov and Sporns, 2010), but most studies have relied on fMRI, which substantially limits the participant groups and numbers that can be studied. High-density EEG recordings offer a comparatively inexpensive easy-to-use alternative, but require channel-level connectivity analysis which currently lacks a common analytic framework and is very limited in spatial resolution. To address this problem, we have developed a new technique for studies of network development that overcomes the spatial constraint and obtains functional networks of cortical areas by using EEG source reconstruction with age-matched average MRI templates (He et al., 1999). In contrast to previously reported channel-level analysis, this approach provides information about the cortical areas most likely to be involved in the network as well as their functional relationship (Babiloni et al., 2005; De Vico Fallani et al., 2007). In this study, we applied source reconstruction with age-matched templates to task-free high-density EEG recordings in typically-developing children between 2 and 6 years of age (O'Reilly, 2012). Graph theory was then applied to the association strengths of 68 cortical regions of interest based on the Desikan-Killiany atlas. We found linear increases of mean node degree, mean clustering coefficient and maximum betweenness centrality between 2 years and 6 years of age. Characteristic path length was negatively correlated with age. The correlation of the network measures with age indicates network development towards more closely integrated networks similar to reports from other imaging modalities (Fair et al., 2008; Power et al., 2010). We also applied eigenvalue decomposition to obtain functional modules (Clayden et al., 2013). Connection strength within these modules did not change with age, and the modules resembled hub networks previously described for MRI (Hagmann et al., 2010; Power et al., 2010). The high temporal resolution of EEG additionally allowed us to distinguish between frequency bands potentially reflecting dynamic coupling between different neural oscillators. Generally, network parameters were similar for networks based on different frequency bands, but frequency band did emerge as a significant factor for clustering coefficient and characteristic path length. In conclusion, the current analysis shows that source reconstruction of high-density EEG recordings with appropriate head models offers a valuable tool for estimating network parameters in studies of brain development. The findings replicate the pattern of closer functional integration over development described for other imaging modalities (Fair et al., 2008; Power et al., 2010).