Automated scoring of sleep in the neonatal lamb.

Abstract

The study of sleep is an important and rapidly expanding area of research that generates large bodies of data. Manual scoring of sleep states from polygraph recordings is a laborious and often subjective task. Even when care is taken, the opportunity for disagreement between investigators and between laboratories remains great. To avoid this difficulty and to reduce the subjectivity of sleep state scoring we have designed a computer-based algorithm for scoring sleep state in the lamb. The algorithm underlying the system relies upon spectral analysis of the electrocorticogram and upon amplitude analysis of the electrooculogram and nuchal electromyogram. Partitioning the spectral power observed within the electrocorticogram (1-4 Hz frequency range) reliably identifies deep quiet sleep. Wakefulness and active sleep are then identified based upon threshold crossings of the electrooculogram and of the electromyogram of the nuchal muscles. We compared the sleep states returned by the algorithm to those scored visually by trained personnel for 1 hour of data collected from each of five 19-day-old lambs. There was good agreement between the two methods of scoring sleep. The percents agreement between the algorithm-derived scores and visual scores were as follows: active wakefulness 97%, quiet wakefulness 87%, quite sleep 85% and phasic active sleep 82%. As such, our algorithm provides a fast, reliable and objective method for scoring sleep state in the young lamb.