Sleep and wake classification with actigraphy and respiratory effort using dynamic warping.

Area of Science:

• Biomedical Engineering
• Sleep Science
• Data Analysis

Background:

• Automatic sleep and wake classification is crucial for sleep disorder diagnosis.
• Current methods using actigraphy and respiratory effort have limitations.
• Novel feature extraction is needed to improve classification accuracy.

Purpose of the Study:

• To propose and evaluate dynamic warping (DW) based features for automatic sleep and wake classification.
• To compare the performance of DW features with existing respiratory and actigraphy-based features.
• To assess the potential of DW features as a non-ECG alternative for sleep analysis.

Main Methods:

• Developed two novel features using dynamic warping (DW) to quantify respiratory effort dissimilarity between sleep and wake states.
• Applied DW in time and frequency domains to align respiratory effort epochs.
• Evaluated DW features using a linear discriminant classifier on a dataset of 15 healthy subjects.

Main Results:

• DW-based features achieved a Cohen's Kappa (κ) of 0.59, outperforming existing respiratory features.
• Combining actigraphy and DW features resulted in κ = 0.66 and 95.7% accuracy.
• The proposed method is comparable to actigraphy and cardiorespiratory approaches but avoids ECG recording.

Conclusions:

• Dynamic warping (DW) offers a significant improvement for sleep and wake classification using respiratory effort.
• DW-based features provide a valuable, non-invasive tool for sleep analysis.
• This approach enhances the accuracy of automatic sleep staging without requiring ECG data.