Slow-wave sleep estimation on a load-cell-installed bed: a non-constrained method

Abstract

Polysomnography (PSG) involves simultaneous and continuous monitoring of relevant normal and abnormal physiological activity during sleep. At present, an electroencephalography-based rule is generally used for classifying sleep stages. However, scoring the PSG record is quite laborious and time consuming. In this paper, movement and cardiac activity were measured unobtrusively by a load-cell-installed bed, and sleep was classified into two stages: slow-wave sleep and non-slow-wave sleep. From the measured cardiac activity, we extracted heartbeat data and calculated heart rate variability parameters: standard deviation of R–R intervals SDNN, low frequency-to-high frequency ratio, alpha of detrended fluctuation analysis and correlation coefficient of R–R interval. The developed system showed a substantial concordance with PSG results when compared using a contingency test. The mean epoch-by-epoch agreement between the proposed method and PSG was 92.5% and Cohen's kappa was 0.62.