• Citer

On-line automatic sleep scoring system involving Bayesian filtering

par Lacroix, B.P.;Hanus, Raymond
Référence Measurement, 2, 3, page (156-160)
Publication Publié, 1984

Article révisé par les pairs

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

In most sleep laboratories, sleep analysis is still routinely performed by a tedious visual examination of electrophysiological signal recordings (3 h per night). In the present paper, we have developed a system for on-line automatic sleep analysis. The method involves three major steps. First, data compression is achieved by extracting outstanding spectral features of the EEG, and a number of phasic events (Spindles, eye movements, etc). In the second step, a local sleep stage decision is evaluated in terms of probability, according to Bayesian theory. Finally, we perform Bayesian smoothing of the hypnogram, introducing the dynamics of sleep physiology, in order to score the sleep stages in a much broader context. The system is running on a PDP 11-44 computer and keeps over 85% reliability when compared with visual analysis. The results have been averaged on 30 sleep records from 10 healthy volunteers aged between 14 and 40. © 1984.