Quantitative Analysis of the Human Sleep Cycle Using Automatic Smoothing Filters

Abstract

Human sleep is divided into rapid eye movement (REM) sleep and non-REM (NREM) sleep, and sleep cycles consisting of NREM sleep followed by REM sleep are a fundamental unit of one-night sleep. Although sleep cycle characteristics may have clinical significance, even qualitative analysis, needless to say quantitative analysis is rarely performed in routine polysomnography (PSG) examination. This is partly because the actual hypnograms have many irregular transitions between stages and appear different from the typical pattern. In order to make the hypnogram simple and the visual judgement of sleep cycle easy, we designed filters to process raw hypnogram data into a simplified schematic one. One group of filters are designed for smoothing the data, converting short intervening protruding epochs into flat continuous stages. The other group of filters bind two similar stages, namely NREM stage1 and 2, and NREM stage3 and 4 into one category, respectively. With these filters, the actual hypnograms are transformed into simplified form, and it became easy for clinicians and patients to perceive the sleep cycle properties. We applied the filters to publicly available sleep stage data and confirmed their validity and efficiency.