[Possibilities and limitations of the automatic analysis of sleep stages using the Oxford system]

34 ambulatory polygraphic sleep-records of young adults were carried out using the Oxford Medilog 9000 recorder. All of the records underwent automatic evaluation by the Oxford Sleep Stager. 10 cases were, in addition, evaluated visually. 4 records could not be analyzed because of artifacts. Thus, 30 subjects were used in the end for establishing standard values of sleep parameters. Automatic versus automatic comparisons and automatic versus visual comparisons of ten polysomnograms were performed epoch-by-epoch. The automatic versus automatic comparison showed agreement in 85.5% of epochs, and the automatic versus visual comparison in 70.4% of epochs. The Sleep Stager showed the best agreement with the visual evaluation in epoch-by-epoch comparison in the REM-stage (93.9%), the worst in stage 1 (32.4%). In calculating the percentages of total night sleep for the respective sleep stages, the automatic analysis showed strongest disagreement with visual evaluation in the REM-stage (plus 7.5%) and in stage 2 (minus 5.2%).