Abstract
Sleep is an essential physiological phenomenon which is regulated by fundamental sleep-wake cycles. Sleep is formed of non-rapid-eye-movement (NREM) and REM stages where NREM and REM sleep stages alternate with wakefulness in a whole night sleep (i.e., typically consists of several sleep-wake cycles). During sleep, brain is active and the activity also alters with changing vigilance states (VS). Furthermore, physiological or external changes in the brain structure might influence brain activity during sleep. Effects of ageing, sex differences, and pharmacological manipulations have been widely investigated in sleep research using Fourier Transform. However, the use of non-linear analysis techniques might be more suitable in analysing non-linear and non-stationary signals (e.g., electroencephalogram (EEG)). Therefore, non-linear analysis has been used within this PhD with the hypothesis that these methods might reveal hidden characteristics in the changing brain signals that are difficult to detect with traditional EEG power spectral density analysis. The use of non-linear analysis techniques (e.g., symbolic dynamic analysis (SDA)) will allow to further dissect the physiological significance of activity-dependent changes of neuronal networks across sleep-wake cycles, as well as the significance of brain activity patterns during waking, sleep, sleep deprivation (SD), or induced by sleep-promoting drugs and pharmacological treatments. In this PhD, rodent and human sleep EEG recordings were analysed using SDA methods: Lempel-Ziv complexity (LZC), Permutation Entropy (PE) and Permutation Lempel-Ziv complexity (PLZC). All the methods were able characterise different VS with wakefulness and REM sleep resulting in higher measures of complexity compared to NREM sleep suggesting an active state of the brain in these VS. This was measured in all datasets assisting the hypothesis on the usefulness of these techniques in sleep research by providing the minimum requirement for sleep analysis. In addition to this, SD significantly reduced complexity in the following sleep period supporting the compensation process for the lost sleep by the increased in slow wave activity which was reflected as reduced complexity in this study. Furthermore, a low dose tiagabine administration’s sleep compensation promoting effect was found in mice. Moreover, ageing was identified as a main effect on changes in brain activity. These changes were more pronounced in the old age where complexity was significantly lower compared to young age. On the one hand, this was found with all three methods and contributing to the hypothesis that these techniques reveal structural dynamic changes due to physiological alterations. On the other hand, no significant differences in complexity across genders were found suggesting the underlying mechanisms to maintain sleep-wake cycles are similar for men and women. This finding with further investigation might corroborate to question the need to use both genders in drug trials. Furthermore, significant changes in brain activity were found at different times of the sleep period highlighting the changes occurring within VS as sleep progresses. This also has an impact on the way sleep stages are scored and investigated which are influenced by different brain activity levels within each VS throughout the entire sleep. All in all, this study achieved to support its hypothesis of determining the changes in brain activity as a complexity measure by characterising sleep under physiological and pharmacologically induced EEG datasets in mice and in humans. The study was a novel application to analyse sleep in these conditions. However, with further analysis performed on larger datasets, its findings together with surrogate data analysis proved SDA techniques’ robust usability which can complement the gold standard FT analysis in sleep research.