Automatic Oxford-Medilog 9200 Sleep Staging Scoring: Comparison with Visual Analysis

We report the results of a comparison study between automatic sleep staging using the ambulatory Oxford-Medilog 9200 system and visual analysis in a group of 30 subjects. Sleep analysis was performed during the second nighttime period of each 48-h polygraphic monitoring. The automatic analysis determined a significant underestimation of total sleep time and stage 2 non-rapid eye movement (NREM) sleep and an overestimation of stage 1. No significant differences were found between the two analyses regarding slow-wave sleep and REM sleep. The differences between automatic and visual analysis of stage 2 were correlated positively with the amplitude and duration of K-complexes and negatively with the amplitude and frequency of spindles. Differences in electroencephalographic characteristics of these sleep figures were correlated negatively with age. Our results showed that (a) the Oxford-Medilog 9200 sleep staging system did not improve the scoring of stages 1 and 2 NREM sleep and (b) the degree of differences between visual and automatic analysis are related to the amplitude and frequency of sleep figures, i.e., spindles and K-complexes, used in the scoring of stage 2.