Sleep staging using cardiorespiratory signals

• Published in: Somnologie : Schlafforschung und Schlafmedizin = Somnology : sleep research and sleep medicine Vol. 11; no. 4; pp. 245 - 256
• Main Authors: Redmond, S. J., de Chazal, P., O'Brien, C., Ryan, S., McNicholas, W. T., Heneghan, C.
• Format: Journal Article
• Language: English
• Published: Heidelberg Springer Nature B.V 01.12.2007
• Subjects: Accuracy; Electrocardiography; REM sleep; Respiration; Sleep apnea
• ISSN: 1432-9123; 1439-054X
• DOI: 10.1007/s11818-007-0314-8

SummaryQuestion of studyWe recently investigated the possibility of obtaining simplified Sleep- Wake-REM sleep stage information from subjects being assessed for Obstructive Sleep Apnea Syndrome (OSAS), using only electrocardiogram and respiration signals. The utility of such a system may be limited somewhat by the presence of OSAS in the patient group (in various degrees of severity). This study examines the effectiveness of such a system when applied to a subject group in which Sleep Disordered Breathing (SDB) is absent.Patients and methodsThe study examined a database of 31 male subjects (Age = 42.0 ± 7.4 years, BMI = 30.7 ± 3.0 kg/m2). There was no significant presence of SDB in any of the subjects (AHI = 1.4 ± 1.2 events/h). A full polysomnography recording was obtained for each subject, including EEG, submental EMG and EOG for sleep staging.An automated sleep-staging system based solely on a single electrocardiogram signal and an inductance plethysmogram estimate of respiratory effort was developed. Features providing useful discrimination of sleep states were derived and the performance of both linear and quadratic discriminant classifiers were compared in correctly labeling 30-second epochs. The use of a time-dependent a priori probability in the classifier models was also investigated.ResultsThe best performance obtained was achieved by a linear discriminant classifier model using a time-dependent a priori probability. For a 3-class (W, S, R) system an agreement of κ = 0.45 was seen,which increases to κ = 0.57 when a simplified 2-class (W, S/R) system is considered. This corresponds to an epoch sleep-wake classification accuracy of 89%.ConclusionsCardiorespiratory signals can provide sleep-wake staging accuracy which is comparable to actigraphy. Classification accuracy is not significantly altered by the presence or absence of sleep disturbed breathing. Cardiorespiratory-based sleep staging may be a useful addition to home sleep apnea monitoring systems.