Figure 1

Computation of recursive dynamic functional connectivity of a triplet of electrodes. (a) The band-pass filtered EEG data of a set of three electrodes is processed using a sliding window to generate the 1st order dynamic connectivity time-series. The same sliding window is applied again on the 1st order dynamic connectivity time-series just obtained to produce the 2nd order dynamic connectivity time-series. The process of recursively applying the sliding window is carried out until the 4th order dynamic connectivity time-series is obtained, which is also the 5th set of time-series when including the original EEG data. In the second step, each of the five sets of time-series are summarized by taking correlation coefficients over their entire duration to compute three static connectivity measures at each order. (b) The three static connectivity measures computed at each order are then assigned to a coordinate value of a three dimensional point representing that order. Finally, all the five points corresponding to the five sets of trivariate time-series are connected with line segments to give a 5-point recursive dynamic connectivity pattern in a 3-dimensional space. (c) The multi-order connectivity of a triplet of electrodes can also be viewed as a hierarchical graph. Each triangle represents a particular order with the edges of the triangle representing the static connectivity measures. For the first order (outermost triangle) the nodes are the three electrodes used for rdFC analysis. The dynamic connectivity of these electrodes is seen as activity of pseudo nodes (dotted circles) that are connected by second order static connectivity measures. As we traverse inwards subsequent higher order static connectivity is represented as edges that connect their respective higher order pseudo nodes.