Automatic sleep apnea detection: analysis of apnea distribution with respect to sleep stages, depending on the severity of sleep apnea

Abstract

Introduction
The present study investigates the distribution of apnea events with respect to sleep stages, depending on the severity of sleep apnea. The results of this study are based on data recorded in the SIESTA project. A new apnea detection software, which was developed recently for the Somnolyzer24x7, will be introduced.

Method
The apnea detection algorithm is based on 4 polysomnographic signals: oxygen saturation, nasal airflow, movement of the chest wall and of the abdomen. Oxygen saturation is resampled to 4 Hz and peaks of this signal are determined in order to extract intervals of oxygen desaturation. Intervals of decreased airflow are calculated by comparing the signal with a smoothed version of itself and are assigned to two classes: an amplitude decrease of more than 50% and more than 80%. From both the chest and the abdominal movement signals, the same intervals are extracted. The detection of apnea events is accomplished by running through a decision tree resulting either in central, obstructive or mixed apneas or hypopneas. Using the automatically determined R&K stages, all apnea events were further assigned to the sleep stages. For the sake of comparability of measurements, the time with apnea events during the different sleep stages was determined (in s) and related to the total time of apneas, resulting in percentages. The polysomnograms of 50 subjects (43 males and 7 females, aged 51+/-10 years) were divided into two groups: 22 subjects with an AHI <= 15 (group 1) and 28 subjects with an AHI > 15 (group 2).

Results
Group 1 (mild apnea) had a mean AHI of 6.8. The time with apnea events in stages 1-4 and REM was (in % of the total time of apneas) 26.4, 48.0, 1.0, 0.6 and 24.0%, respectively. The corresponding sleep stages (in % of the total sleep time) were 14.7, 52.6, 8.4, 4.0 and 20.3%, respectively. Group 2 (moderate to severe apnea) had a mean AHI of 31.2. The time with apnea events in stages 1-4 and REM was 23.2, 55.2, 3.3, 0.6, and 17.8%, respectively. The corresponding sleep stages were 19.5, 52.9, 5.8, 2.9 and 18.8%, respectively. Whereas t-tests revealed no significant difference in the distribution of sleep stages between the two groups, a significant difference for the occurrence of sleep apnea was found in stage 3.

Conclusion
Patients with moderate to severe apnea spent significantly more time with apnea in stage 3 than patients with mild apnea, even after correction for the different total time with apnea in both groups. However, our data do not suggest a clear relationship between the distribution of apnea/hypopnea events in the different sleep stages and the severity of the disease.


Go back