Transient Topographical Dynamics of the Electroencephalogram Predict Brain Connectivity and Behavioural Responsiveness During Drowsiness

Abstract

As we fall sleep, our brain traverses a series of gradual changes at physiological, behavioural and cognitive levels, which are not yet fully understood. The loss of responsiveness is a critical event in the transition from wakefulness to sleep. Here we seek to understand the electrophysiological signatures that reflect the loss of capacity to respond to external stimuli during drowsiness using two complementary methods: spectral connectivity and EEG microstates. Furthermore, we integrate these two methods for the first time by investigating the connectivity patterns captured during individual microstate lifetimes. While participants performed an auditory semantic classification task, we allowed them to become drowsy and unresponsive. As they stopped responding to the stimuli, we report the breakdown of alpha networks and the emergence of theta connectivity. Further, we show that the temporal dynamics of all canonical EEG microstates slow down during unresponsiveness. We identify a specific microstate (D) whose occurrence and duration are prominently increased during this period. Employing machine learning, we show that the temporal properties of microstate D, particularly its prolonged duration, predicts the response likelihood to individual stimuli. Finally, we find a novel relationship between microstates and brain networks as we show that microstate D uniquely indexes significantly stronger theta connectivity during unresponsiveness. Our findings demonstrate that the transition to unconsciousness is not linear, but rather consists of an interplay between transient brain networks reflecting different degrees of sleep depth.

Keywords: Brain connectivity; Drowsiness; EEG microstates; Responsiveness.

Figures

Fig. 1

Reaction times and data segmentation into responsiveness and unresponsiveness for individual participants. The horizontal axis represents recording time and the vertical axis represents reaction time in seconds. Blue markers indicate responses, while orange markers indicate misses. The blue area corresponds to the 5-min period of responsiveness, while the orange area corresponds to the 5-min period of unresponsiveness

Fig. 2

Spectral power topography and WPLI frontoparietal connectivity at alpha (9.5–10.5 Hz) and theta (5–6 Hz) peaks before and after the loss of responsiveness. Values are averaged across participants. With loss of responsiveness, power over parietal sensors and connectivity between frontal and parietal sensors shifted from the alpha to the theta band

Fig. 3

Microstate topographies computed across all subjects. These topographies are plotted in correspondence with the four canonical microstate topographies commonly described in literature. Microstate topographies reported by Brodbeck et al. (2012a, b) and Koenig et al. (2002a, b) are shown for comparison (reproduced here with permission)

Fig. 4

Microstate parameters before and after the loss of responsiveness in drowsiness. Within each panel of grouped scatter box plots, inner boxes represent the standard error of the mean for each microstate parameter, and outer boxes represent the standard deviation. The mean is shown by a continuous line, the median is shown by a dotted line, and individual participants are shown as dots. Asterisks show a significant within-subject main effect of state for a microstate. Duration, temporal coverage and GEV of microstate D all significantly increased during unresponsiveness

Fig. 5

Classification performance, computed as the area under the ROC curve, for a support-vector machine (SVM) trained using 5 s of pre-stimulus data to classify responses and misses. Input features to the classifier were microstate parameters or the theta–alpha ratio, individually or combined. Within each group of grouped scatter box plots, inner boxes represent the standard error of the mean, outer boxes represent the standard deviation. The mean is shown by a yellow line, the median is shown by a green line (where distinct from the mean), and individual participants are shown as dots. Microstate parameters were able to predict responsiveness at an individual trial level across subjects, with a performance similar to that of the theta-alpha ratio

Fig. 6

Spectral power (a) and WPLI connectivity (b) captured during individual microstates before and after loss of responsiveness due to drowsiness. Channel-wise relative power at each frequency bin was calculated as the power at that bin as a percentage of total power within 1–20 Hz. Within each subject, for both relative power and WPLI connectivity, the median across channels is plotted. Figures show the grand average over all subjects. c Shows the effect size, computed as Cohen’s d, of the interaction between behavioural state and microstate at each frequency bin for power contributions and for connectivity. By convention, 0.2, 0.5 and 0.8 denote small, medium and large effect sizes, respectively. The interaction between microstate and behavioural state was stronger in connectivity than in power

Fig. 7

Frontal and frontoparietal WPLI connectivity at theta peak (5–6 Hz). Microstate D captured significantly higher theta connectivity within frontal and between frontoparietal sensors during unresponsiveness, compared to microstates A–C