Related Experiment Videos
Cardiorespiratory-based sleep staging in subjects with obstructive sleep apnea
Stephen J Redmond1, Conor Heneghan
1Department of Electronic Engineering, University College Dublin, Belfield D4, Ireland. Stephen.Redmond@ee.ucd.ie
IEEE Transactions on Bio-Medical Engineering
|March 15, 2006
Summary
This study developed a cardiorespiratory-based sleep staging system, achieving moderate accuracy (79%) but highlighting significant subject-specific performance variations. Subject-independent systems showed lower accuracy, indicating limitations for general use.
Area of Science:
- Biomedical Engineering
- Sleep Medicine
- Signal Processing
Background:
- Automatic sleep staging is crucial for diagnosing sleep disorders.
- Cardiorespiratory signals offer a non-invasive alternative to traditional polysomnography.
- Previous systems often rely on electroencephalography (EEG), which can be cumbersome.
Purpose of the Study:
- To develop and evaluate a cardiorespiratory-based automatic sleep staging system.
- To assess the performance of subject-specific versus subject-independent classifiers.
- To investigate the influence of sleep-disordered breathing severity (Apnea-Hypopnea Index) on system accuracy.
Main Methods:
- A three-state sleep model (Wakefulness, REM, non-REM) was implemented.
- Features extracted from RR-intervals, respiratory effort (inductance plethysmography), and electrocardiogram-derived respiration (EDR) were used.
- Quadratic discriminant classifiers were trained and tested using subject-specific and subject-independent data subsets.
Main Results:
- The subject-specific cardiorespiratory classifier achieved 79% accuracy (kappa=0.56).
- Subject-independent classification accuracy dropped to 67% (kappa=0.32).
- Subject-specific performance was better for low Apnea-Hypopnea Index (AHI) subjects; EEG-based classifiers showed higher, more robust accuracy.
Conclusions:
- Cardiorespiratory signals provide moderate sleep-staging accuracy.
- Significant subject dependence limits the generalizability of cardiorespiratory-based systems.
- EEG-based methods remain superior for robust, subject-independent sleep staging.