Structural EEG signal analysis for sleep apnea classification

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Diagnosing sleep apnea is crucial for preventing associated health complications.
• Electroencephalogram (EEG) signals offer a potential avenue for non-invasive sleep apnea assessment.
• Analyzing transient signal changes during sleep apnea events requires robust signal processing techniques.

Purpose of the Study:

• To develop a sleep apnea scoring approach by comparing parametric and non-parametric Power Spectral Density (PSD) estimation methods.
• To analyze transient EEG signals from different brain regions (C4-M1 and O2-M1) for sleep apnea detection.
• To evaluate the efficacy of various PSD methods in identifying sleep apnea transition states.

Main Methods:

• Employed several Power Spectral Density (PSD) estimation methods: Burg, Yule-Walker, periodogram, Welch, and multi-taper.
• Examined the detection of sleep apnea transition states (pre-apnea, intra-apnea, post-apnea) using these PSD methods.
• Utilized statistical analysis and K-nearest neighbor (KNN) classification to differentiate between methods and brain regions.

Main Results:

• Statistically significant differences were observed between parametric and non-parametric PSD methods in analyzing delta, theta, alpha, and beta EEG bands during apnea transitions.
• EEG signals recorded from C4-M1 and O2-M1 brain regions exhibited statistically different PSD characteristics.
• KNN classification confirmed the distinctiveness of PSD patterns derived from different brain regions and PSD estimation techniques.

Conclusions:

• The choice of PSD estimation method significantly impacts the statistical analysis of sleep apnea transition states.
• EEG signals from different brain regions (C4-M1 vs. O2-M1) provide distinct information relevant to sleep apnea detection.
• Combining varied PSD methods with analysis of specific brain regions enhances the potential for accurate sleep apnea scoring.