Lempel-Ziv complexity

Lempel-Ziv Complexity (LZC) was computed as a surrogate of complexity to reflect the spatiotemporal repertoire across scalp potentials. LZC is a method of symbolic sequence analysis that measures the complexity of finite length sequences (Lempel and Ziv, 1976), which has been shown to be a valuable tool to investigate brain states related to consciousness and cognition (Casali et al., 2013; Abásolo et al., 2015; Schartner et al., 2015; Hudetz et al., 2016; Schartner et al., 2017). The calculation of LZC requires a binarization of the multichannel EEG data. In this study, we used the implementation as described in Schartner et al., 2015; Schartner et al., 2017, and calculated the instantaneous amplitude from the Hilbert transformed EEG signal for each channel, which was binarized using its mean value as the threshold for the current channel (supplementary analysis was performed to test the effect of threshold selection). The data segment was then converted into a binary matrix, in which rows represent channels and columns represent time points. LZC was computed by searching the spatiotemporal matrix time point by time point and counting the number of different spatial patterns across different time points. Thus, the resultant measure captures the complexity or diversity in both temporal and spatial domains.

For implementation, the average signal was first subtracted from all channels in order to remove the effect of common reference, and then the multichannel EEG epochs were divided into non-overlapping 4 s windows to compute the LZC, with the resultant LZC values being averaged across all the windows for each studied epoch. In line with previous studies, we normalized the original LZC by the mean of the LZC values from N = 50 surrogate data sets generated by randomly shuffling each row of the binary matrix, which is maximal for a binary sequence of fixed length (Schartner et al., 2015; Schartner et al., 2017) (supplementary analysis was performed to test alternative methods in the generation of surrogate data).