Estimating sleep time from non-EEG-based PSG signals in the diagnosis of sleep-disordered breathing.

Abstract

The purpose of this study was to use non-EEG PSG signals to estimate TST in order to diagnose SDB with a greater sensitivity than type 3 device methodology that relies on TRT. Movement patterns were obtained from the thoracoabdominal signals of adult PSG recordings (n=60) in the laboratory and the home. Parameters obtained allowed, with 95% certainty, identification of sleep and wake based on the duration of movements and quiescent time (Qd). Snoring, apneas, and hypopneas indicated sleep with 100% certainty. The method was tested in a different set of PSG recordings (n=80). Subjects lay awake and immobile for longer in the laboratory (QdLAB=27.4 (12.1, 62.0), QdHOME=16.0s (8.0, 36.0); p<0.0001) but asleep and immobile for longer at home (QdLAB=65.2 (23.0, 121.4), QdHOME=95.0s (44.5, 247.5); 0.005). Only 5% of wake Qd periods were >173s in the laboratory and >105s at home. In both locations, 95% of movements during sleep were <10s. Experimental TST values were 21min shorter than EEG-defined TST and, combined with fewer scored respiratory events, produced AHI values that were 1.6 events/h lower than the reference. The experimental TST increased the sensitivity of SDB diagnosis from 73 to 97%. In the sleep laboratory, subjects are immobile for longer periods when awake and for shorter periods when asleep. The experimental TST was similar to EEG-defined TST and could be used to diagnose SDB with a much higher sensitivity than the type 3 method.