The detection of drowsiness and sleep onset periods from ambulatory recorded polygraphic data

A 30 min sample recording at the sleep onset of 7 healthy male subjects was used to further develop a computer-scoring algorithm applied earlier for the analysis of MSLT recordings. The performance of this algorithm was tested on 7 patients with obstructive sleep apnea by using 6 h daytime recordings including drowsiness and sleep episodes. The total epoch-by-epoch agreement between visual and computer scoring was over 90% and the accurate detection rate of non-REM sleep was 64%. The hypnograms produced by the computer scoring corresponded sufficiently to those obtained by visual scoring. Our automatic scoring system can give a good estimation of the daytime vigilance profile but for clinical diagnosis the results have to be verified visually. However, by using modern digital recording, analyzing and scoring techniques the speed of analysis and thus the costs can markedly been reduced as compared to traditional visual analysis.