Automatic Sleep Stage Classification Using Electro-oculography

In this thesis automatic sleep stage classification was developed and evaluated. The method was based on signals recorded by electro-oculography electrodes. Monitoring sleep is important for the diagnosis of sleep disorders. Altered sleep is related to obesity and diabetes, and loss of sleep may lead to daytime sleepiness which in turn may cause accidents. Standard sleep stage measurement requires the application of multiple electrodes by trained professionals. Signals are then classified visually in a timeconsuming and subjective process. Many automatic sleep classification methods also exist. Some methods work with self-applicable, usually forehead, electrodes. However, the use of standard sleep electro-oculography electrode placement enables the recording of frontal EEG, EMG and EOG using a single electrode pair. Nearly 300 sleep recordings were used to develop automatic methods for separating wakefulness and sleep stages during intentional night-time sleep and during unintentional daytime sleep through maintenance of wakefulness tests (MWT). Signals detected using only standard electro-oculography electrodes were used for automatic sleep stage classification. The signals were recorded both with and without the mastoid reference electrode. Results were also compared with activity-based methods, and for reference, we also recorded EEG and submental EMG tonus. Reference sleep stage scoring was carried out visually according to the Rechtschaffen and Kales standard. Reasonable sleep stage information could be obtained using self-applicable electrooculography electrodes combined with automatic analysis. This developed self-applicable automatic sleep staging system would make large scale ambulatory sleep studies plausible for screening sleep disorders and investigating the relationship between irregular sleep and health