engleski

Pitfall of automatic sleep stage detection

The major challenge in polysomnographic (PSG) analysis is finding the compromise between the speed and the accuracy of determination of sleep parameters. All laboratory PSG acquisition software incorporates algorithms for some level of automated scoring. However, for sleep stage scoring at least, the implemented set of algorithms is not accurate enough, and data are usually additionally scored by sleep technicians. In this study, we discuss results of automated stage scoring, generated with commercial software for Alice4 (Respironics, USA), and MEPAL (MAP, Germany) devices, on visually- scored test databases. Test databases contain 30s epochs of PSG signal from S1, S2, SWS and REM sleep stages, derived from 20 healthy male subjects, as well as epochs from 18 male patients with obstructive sleep apnea. All epochs are equally scored by 2 independent experts. Sensitivity and specificity of automatic detection for particular sleep stage and for particular subject group is discussed.

sleep stage detection

nije evidentirano