The most relevant limitation of the present study is the selection of a relatively small number of ROIs (i.e. 12 ROIs in total, six for each subject) to obtain hyper-brain networks. This aspect limits the power of the graph theoretical approach and restricts the cortical networks to a subset of predefined ROIs. This limitation is mainly due to the Partial Directed Coherence index (PDC), which is based on the generation of a valid multivariate autoregressive model (MVAR) from the estimated cortical time series. The PDC is one of the most powerful methods to reveal directed information flows between time series. Other simple methods, like spectral coherence [25], are currently available in the literature. Those bivariate methods allow the estimation of the functional connectivity among a larger number of cortical signals. Nevertheless, the advantage of MVAR models with respect to other standard bivariate methods have been already demonstrated [26]. In fact, they can efficiently detect and remove the statistical spurious links from the functional connectivity estimation, even if in most cases, like PDC, MVAR methods need a large amount of data
