Let's get digital: Quantifying sleep EEG microarchitecture from polysomnography

Abstract

Polysomnography (PSG) is an objective assessment used to diagnose sleep disorders and includes recording brain activity during sleep using electroencephalography (EEG). Conventional qualitative analysis of the EEG signal to derive sleep macroarchitecture measures such as sleep stages provide little prognostic value for patients with disrupted sleep. Using modern computing, detailed analysis of brain activity using quantitative EEG (qEEG) techniques can automatically quantify key sleep EEG oscillations, or microarchitecture events, such as slow waves and sleep spindles. These brain waves play a pivotal role in sleep stability, sleep-dependant cognitive processes, and are known to change with age and disease. My thesis explores the potential utility of qEEG analysis to understand how individual variability in sleep neurophysiology may influence patient daytime functioning in three sleep disordered populations. It also aims to contribute to further understanding the relationship between sleep and cognition in those with disordered sleep. In project 1, I extracted 20 minute segments of EEG power spectral analysis (10 minutes either side of sleep onset) in 94 patients with insomnia disorder. The short-sleep duration with difficulty initiating sleep ‘phenotype’ had significantly reduced EEG spectral power across the scalp and worse neurocognitive test performance compared to insomnia patients without sleep onset difficulties. In the second project, I used quantitative EEG (qEEG) analysis to investigate sleep EEG microarchitecture correlates of neurobehavioural performance after 24-h of extended wakefulness in untreated obstructive sleep apnea (OSA). During baseline PSG, greater central EEG slowing in REM and lower frontal spindle density in NREM sleep was associated with worse neurobehavioural performance. In the third project I used a case control study and found qEEG measures of slow wave and spindle activity to be significantly reduced across the scalp in 29 Parkinson’s disease (PD) compared to 14 healthy controls, and were associated with worse performance on tasks of executive function. Digital recording is ubiquitous in contemporary clinical PSG, however traditional analysis methods designed to summarise paper recordings are still used exclusively, resulting in lost physiological detail pertinent to sleep and daytime function. Slow waves and sleep spindle analysis using qEEG could help explain individual variation in vulnerability to the negative consequences of sleep loss, and may provide a prospective biomarker of cognitive decline. Using qEEG to ‘phenotype’ individuals might also offer an avenue for personalised treatment approaches.