INTRODUCTION TO DEEP SLEEP STAGE MANUAL

These studies are then evaluated by experts for the presence of events of clinical relevance, as determined by standards created by the American Academy of Sleep Medicine (AASM), such as the number of blood oxygen desaturations, micro-arousals, leg movements, periods of cessated breathing, etc. Briefly, a standard clinical sleep study consists of a full-night polysomnography (PSG) comprising electroencephalography (EEG), electrooculography (EOG), electromyography (EMG), electrocardiography (ECG), thoraco-abdominal inductance plethysmography, oronasal thermal flow, nasal pressure, and blood saturation recordings. Sleep staging is important to the analysis of human sleep with about 845,000 sleep studies performed in 2014 in the United States alone.

Future researchers should take multiple datasets into account when developing sleep scoring models. We found that using data from many different sources improves classification performance, and that models trained on single-origin data generalize inconsistently to new data. Our algorithm was tested under different conditions to compare generalizability. We developed an algorithm for automatic scoring of sleep stages using raw polysomnography data and obtain state-of-the-art classification performance on a large number of test subjects. Most research in automatic sleep stage classification models use small-scale data from a single origin, and it is unknown how these models generalize to new data.

INTRODUCTION TO DEEP SLEEP STAGE MANUAL

Manual annotation of polysomnography studies is subject to human bias with multiple studies showing variations in how sleep experts score sleep.

Introduction to deep sleep stage