Functional and Effective Connectivity in Healthy Human EEG across Sleep Stages.

• ¹ Erasme Academic Hospital - ULB, Department of Psychiatry, Belgium
• ² Faculty of Psychology and Pedagogical Sciences, University of Gent, Department of Data Analysis, Belgium

Cerebral activity during sleep can be approached by the measurement of functional and effective connectivity in healthy young men. Functional connectivity (FC) reflects the statistical dependencies among even remote brain regions by using coherence analysis (COH) and generalized partial directed coherence (GPDC), while effective connectivity (EC) investigates the directional pathways of information flow across brain regions by applying Granger Causality (GC).
We analyzed the first 20 epochs of each sleep stages (wake, sleep stages 1 to 4 and REM sleep) of 9 healthy young men. Sleep stages were visually scored in accordance with the Rechtschaffen and Kales criteria. All subjects did not suffer from somatic, psychiatric or sleep disorders and did not take any medication. In regard of specific sleep EEG frequencies as delta, theta, alpha, sigma and beta, coherence analysis and generalized partial directed coherence, which allows to detect directional influences, were applied between each electrode and each other of the 19 EEG electrodes. Connectivity matrices between the 19 EEG electrodes were also obtained for each sleep stage using multivariate Granger Causality.
Our results demontsrated that coherence analysis showed high values within the same hemisphere during wake and sleep stage 1. During sleep stages 2, 3 and 4 as well as during REM sleep, high correlations were extended to more regions, especially for slow oscillations. Generalized partial directed coherence showed increased interdependencies in each NREM sleep stages in comparison to wake and REM sleep. Again, this pattern was more evident for slow rhythms. Slow rhythms were those for which more differences were observed across all sleep stages. Furthermore, wake and early NREM sleep stages were those for which the differences between bands were more pronounced. Using Granger Causality, during wake, hetero-hemispherical activity was observed reciprocally in frontal, occipital and occipito-parietal regions, as well as long-range connections. Across sleep stages 1 to 4, the information flow decreased in the posterior brain regions. During REM sleep, connectivity was limited to short range links in occipital and fronto-temporal regions, predominantly on the left side.
In conclusion,the directed and undirected coherence (COH and GPDC) measurements as well as the effective connectivity (GC) proved efficient and robust in mapping the three main sleep states (Wake, NREM and REM sleep). Further analyses will investigate connectivity in the transition between sleep stages and in clinical populations.