EEG connectivity measures wPLI and wSMI identify distinctive differences in brain functional interactions during wakefulness and sleep

Objectives/Introduction: Functional connectivity measures represent a powerful tool to investigate brain inter-regional interactions in distinct behavioural states. We recently demonstrated that the weighted Phase Lag Index (wPLI) and the weighted Symbolic Mutual Information (wSMI), two connectivity measures that are relatively immune to volume-conduction, present a different sensitivity for linear and non-linear interaction dynamics, respectively. Here we investigated whether these two methods may unveil distinct functional differences between wakefulness and sleep, as well as across the distinct sleep stages. Methods: Twelve healthy volunteers (25±4 years, 6F) participated in the study. All subjects underwent hd-EEG recordings (256 channels, 500Hz) during the night (11.30 PM-7 AM) and while awake in bed after sleep (8AM; 6 min, eyes-closed). After pre-processing, 70 2 s data segments (minimum common number across subjects and stages) were extracted from wake data (W) and from the first sleep epochs of stable N2, N3 and REM-sleep. For each electrode, the median connectivity between each pair of channels was computed in delta, sigma and gamma bands. Friedman tests were used to investigate stage-related effects, while signed-rank tests were used in posthoc comparisons (p< 0.05, Bonferroni correction). Results: Delta-wPLI was significantly increased in all sleep stages relative to W, whereas delta-wSMI was decreased in REM-sleep as compared to W and N2 (p< 0.05; N = 12). In sigma-band, wPLI was significantly higher in N2 relative to W, while wSMI was significantly higher in N2 and N3 relative to REM-sleep. Finally, gamma-wSMI was significantly lower in REM-sleep relative to N2 and N3, and in W relative to N3. No significant stage-effects were found for gamma-wPLI. Conclusions: All sleep stages were characterized by a significant increase in linear dynamics within the delta range, likely due to the presence of travelling low-frequency oscillations. Sleep spindles may account for changes in sigma-connectivity specific to NREM sleep, while differences in high-frequency activity are consistent with previous work reporting the occurrence of hypercorrelated gamma activity in this stage. These results demonstrate that sleep is associated with both general and stage-specific changes in linear and nonlinearbrain dynamics. Their combined study may offer a valuable tool for assessing changes in level of consciousness under physiological and pathological conditions.