Slow oscillations in the cat, and rodent under various anaesthetics and in natural slow wave sleep (SWS) non­rapid eye movement (non­REM) stage exhibit rhythmically recurring phases of wide spread cortical hyperpolarizing and depolarizing currents. The hyperpolarizing phase (down­state) appears to be a negative deflection and the depolarizing phase (up­state) as a positive deflection in the surface recorded electroencephalogram (EEG), with polarity inversion in the depth of the cortex. Similar patterns were found in the human sleep surface EEG during non­REM sleep, which appeared as propagating waves. In the up­state in animals, the majority of cortical neurons is depolarized and fire action potentials, whereas in the down­state most of the neurons are relatively hyperpolarized and their action potential generating activity is highly decreased. The functional significance of slow oscillation lies in the long term preservation of memory traces, and is also believed to have strong linkages to the generation of epileptic activity.

To elucidate the intracortical neuronal mechanisms of the sleep slow oscillations in humans, laminar multielectrodes were chronically implanted into the cortex of patients with drug resistant epilepsy undergoing cortical mapping for seizure focus identification. Intracortical laminar local field potentials, current source density (CSD), multiple and single unit activity (MUA, SUA) was recorded during quiet slow wave sleep, non­rapid eye movement periods.

Our work was built up of two parts: software engineering that enabled us to process recorded data, and biological analysis to reveal the functioning of the brain during slow wave sleep in human and the underlying physiology. Our main aim was to show that slow oscillation in humans is essentially similar to the animal models. There are no reports up to the present date on human SUA, MUA and CSD data related to slow oscillation, so our work is essential in proving, that similar mechanisms are in effect in humans and in animals during slow sleep oscillations.

The methods and software necessary to examine slow oscillations were successfully developed. We applied previously published, and also developed our own state detection methods. One of our main approach was filtering the signal in a low­frequency range (0.5­3 Hz), then applying a Hilbert transform. Thus we can determine the maximum and minimum points of the signal, corresponding to the up­ and down­states. A different method was applied on multiunit (MUA) recordings. A threshold was set between the baseline level (no action potentials – down­state) and the amplitude level of action potentials (up­state). Action potentials in MUA recordings appear at the beginning of up­states, while they develop on field potential recordings with a certain latency, so this approach produces a much more precise detection in time than other methods.

We used state detection methods to analyze the behaviour of cortical neurons during slow oscillation in deep, non­REM sleep. During depolarization periods (up­state) we observed positive field potentials on the surface and negative potentials in deeper layers, fast (gamma) oscillations, CSD sink in the middle layers and increased firing rate. During hyperpolarization periods (down-state) negative field potentials appeared on the surface and positive potentials deeper, current sources were present in the middle layers, there were no fast oscillations, and cortical neurons remained silent, generated no action potentials.

Our results show for the first time, that the behaviour of slow oscillation in human is similar to cat and other animal models.