Nonstationarities and sleep disturbance detection

We present recent results on sleep apnea-hypopnea quantification. Sleep apnea is
the most common sleep disturbance and it is an important risk factor for
cardiovascular disorders. The diagnostic is obtained trough combined exams,
and we aim to offer an alternative procedure. A sleep apnea event is defined
as a break in the airflow that lasts at least 10 secs. If the air flow is
less than 50\% of normal, the resulting airflow limitation is called hypopnea.
Individuals who suffer from this kind of disorder usually present daytime
sleepiness, loud snoring and restless sleep.
%
Blood pressure, heart rate variability, respiratory variability, and other
cardiorespiratory data could be useful to detect sleep disturbances, and it is
important to emphasize that cardiorespiratory time series are highly
nonstationary, which restricts the use of standard tools of time series
analysis.
%
Hence, we apply a nonparametric segmentation procedure to yield patches where
stationarity is verified. Within each of these locally stationary data
segments, the statistical moments of the signal, such as mean and variance,
remain constant. Segmentation also provides the intrinsic time scales, through
the duration of segment lengths. We show that the occurrence of sleep apnea
events can be quantified by means of blood pressure time series analysis,
and by comparing local quantities to an apnea score previously obtained by
polysomnographic exams, we propose an apnea quantifier based on blood pressure
signal with an accuracy of $82\%$.

Ref: S. Camargo, M. Riedl, C. Anteneodo, J. Kurths, T. Penzel, N. Wessel, Sleep apnea-hypopnea quantification by cardiovascular data analysis, Plos One {\bf 9}, e107581