Theta waves in children's waking electroencephalogram resemble local aspects of sleep during wakefulness

Abstract

Vyazovskiy and colleagues found in rats' multi-unit recordings brief periods of silence (off-states) in local populations of cortical neurons during wakefulness which closely resembled the characteristic off-states during sleep. These off-states became more global and frequent with increasing sleep pressure and were associated with the well-known increase of theta activity under sleep deprivation in the surface EEG. Moreover, the occurrence of such off-states was related to impaired performance. While these animal experiments were based on intracranial recordings, we aimed to explore whether the human surface EEG may also provide evidence for such a local sleep-like intrusion during wakefulness. Thus, we analysed high-density wake EEG recordings during an auditory attention task in the morning and evening in 12 children. We found that, theta waves became more widespread in the evening and the occurrence of widespread theta waves was associated with slower reaction times in the attention task. These results indicate that widespread theta events measured on the scalp might be markers of local sleep in humans. Moreover, such markers of local sleep, seem to be related to the well described performance decline under high sleep pressure.

Figure 1

Representative examples of amplitude differences in wake EEG of a child and an adult. (a) Four seconds of the unfiltered wake EEG of an adult in the evening (male subject, 19.4 years old, from the data set presented in ref. 45). (b) Four seconds of the unfiltered wake EEG of a child in the evening (male subject, 12.6 years old). Note, the amplitude of the wake EEG is much higher in children compared to adults. (c) Example of the theta event detection algorithm: the same 4 seconds EEG as in b are displayed after filtering (Chebyshev Type 2 Filter: pass-band 5 and 9 Hz, stopband: below 4 and above 12 Hz). Grey asterisks mark the theta events detected by the algorithm.

Figure 2

Relationship between detection window size and cluster size. (a) Mean cluster size of theta events for different detection windows, based on an amplitude detection threshold matched to each EEG recording (mean ± SEM of cluster sizes for each detection window in the morning and evening are shown). The larger the detection window, the more channels are involved in the cluster sizes (linear mixed effect model: F_{detection window} = 546.92, p < 0.001, n = 12). No difference in cluster size of theta events was found between morning and evening (linear mixed effect model: F_time = 0.01, p = 0.93 n = 12). (b) The difference (percentage) of mean cluster sizes (i.e. the number of channels involved in a cluster) between different detection windows. Morning and evening are pooled. X-axis represents upper detection window (i.e. 40 = cluster size_{Detection window 40}/cluster size _{Detection window 20}*100). A detection window of 60ms (indicated by the grey dotted line) was selected for analysis.

Figure 3

Definition of widespread theta events based on a detection window of 60 ms. (a) Relationship between amplitude and cluster size in the morning and evening (mean ± SEM of the amplitude for each cluster size are presented for morning and evening. For each subject, mean amplitude was calculated when at least 5 theta events for the given cluster size were detected). Larger amplitudes were associated with increased cluster sizes (linear mixed effect model F_{cluster size} = 32.61, p < 0.001, n = 12). For any cluster size a larger amplitude was detected in the evening compared to morning (linear mixed effect model: F_time = 207.01, p < 0.001, n = 12). (b) 5-percentile bins of cluster sizes (morning and evening pooled). The 85^th percentile corresponds to a cluster size of 20 channels (grey dotted line). (c) Difference in the number of channels involved in a cluster size between each 5-percentile bin from Figure b. Numbers on the x-axis indicate the lower bin (i.e. 5 corresponds to the difference between the 5^th and the 10^th bin). Because the number of channels involved in the cluster size are increasing from the 85th percentile, the 85th percentile (corresponding to a cluster size of 20 channels, see b) was used for cluster size cutoff definition for widespread theta events.

Figure 4

Increase of widespread theta events form morning to evening. (a) Percentage of widespread theta events in the morning and evening (bars represent mean ± SEM, circles represent individual subjects). Percentage of widespread theta events increase in all subject from morning to evening (p < 0.001, paired Student’s T-Test, n = 12). (b) Topographical distribution of widespread theta events changes from morning to evening (evening-morning/morning*100). Percentage of widespread theta events increases globally from morning to evening, which is most pronounced over central and frontal areas (white dotes p < 0.05, paired Student’s T-Test, n = 12, after nonparametric cluster-based statistical testing).

Figure 5

Relationship between widespread theta events and reaction time. Data from morning and evening pooled. (a) Mean reaction time with no widespread theta event (=Without) and with at least one widespread theta event (=With). Reaction times associated with widespread theta are slower compared to reaction times without widespread theta events (mean ± SEM, p = 0.036, paired Student’s T-Test, n = 18). (b) Topographical comparison of widespread theta events and reaction time. All reaction times were split into two groups: the fast reaction time (defined as the fastest 40% reaction times) and the slow reaction time group (the slowest 40% reaction times, including the missed stimuli). For each group it was evaluated how often a channel was involved on average in a widespread theta event. A cluster of 11 channels (white dotes) was found, which were significantly more involved in widespread theta events in the slow reaction time compared to the fast reaction time group (white dotes p < 0.05, paired Student’s T-Test, n = 18, after nonparametric cluster-based statistical testing).