Capacitively-Coupled ECG and Respiration for Sleep-Wake Prediction and Risk Detection in Sleep Apnea Patients

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Signal Processing

Background:

• Obstructive sleep apnea (OSA) diagnosis typically requires polysomnography (PSG), limiting comfortable home-based monitoring.
• Accurate detection of wakefulness is crucial for effective sleep analysis in OSA patients.
• Unobtrusive sensors for sleep monitoring at home are needed to improve patient comfort and compliance.

Purpose of the Study:

• To assess the quality of unobtrusive capacitively-coupled electrocardiogram (ccECG) and bioimpedance (ccBioZ) signals during sleep.
• To develop and evaluate a system for sleep-wake prediction using ccECG and ccBioZ signals.
• To identify high-risk OSA patients using the developed sleep-wake prediction framework.

Main Methods:

• Collected simultaneous PSG, ccECG, and ccBioZ data from 187 suspected OSA patients.
• Utilized signal quality indicators (SQIs) for data coverage assessment.
• Employed a multimodal convolutional neural network (CNN) for sleep-wake prediction, adapted to a unimodal network using data augmentation, and derived indices for OSA risk detection.

Main Results:

• Significantly improved ccBioZ signal coverage through acquisition system adaptation.
• Achieved sleep-wake prediction with Cohen's kappa (κ) of 0.39 using PSG respiration and κ = 0.23 using ccBioZ.
• Identified severe OSA patients with κ = 0.61 for PSG respiration and κ = 0.39 using ccBioZ (80.6% accuracy).

Conclusions:

• This study demonstrates the feasibility of sleep-wake staging using capacitively-coupled respiratory signals in suspected OSA patients.
• The proposed framework effectively detects high-risk OSA patients using ccBioZ signals, offering a promising approach for home diagnosis.
• The developed technology and framework hold potential for multi-night follow-up monitoring of OSA patients.