Combining Heart Rate Variability and Oximetry to Improve Apneic Event Screening in Non-Desaturating Patients

Area of Science:

• Biomedical Engineering
• Medical Signal Processing
• Sleep Medicine

Background:

• Obstructive Sleep Apnea (OSA) is a prevalent public health issue with significant health implications.
• Accurate detection of sleep apnea events, particularly in non-desaturating patients, remains a challenge.
• Existing methods may not fully capture the complexities of sleep apnea, especially in diverse patient populations.

Purpose of the Study:

• To develop and evaluate a system for detecting sleep apnea events, focusing on non-desaturating patients.
• To assess the effectiveness of combining Heart-Rate Variability (HRV) and SpO2 features for sleep apnea characterization.
• To compare the performance of the proposed system on different datasets, including Physionet and HuGCDN2014-OXI.

Main Methods:

• Utilized the HuGCDN2014-OXI and Physionet Apnea datasets, encompassing both desaturating and non-desaturating patients.
• Extracted HRV features (spectral, cepstral, nonlinear: DFA, RQA) and SpO2 features (temporal, spectral).
• Employed a Linear Discriminant Analysis (LDA) classifier to analyze individual and combined feature sets.

Main Results:

• Achieved high performance on the Physionet dataset (96.19% success rate, 0.99 AUC) and HuGCDN2014-OXI (87.32% success rate, 0.934 AUC) for apneic event detection.
• Demonstrated superior performance for global OSA diagnosis on HuGCDN2014-OXI (95.74% success rate, 100% sensitivity, 89.47% specificity).
• The combination of HRV and SpO2 features yielded the most accurate results, especially for non-desaturating patterns.

Conclusions:

• Combining HRV and SpO2 features significantly improves the accuracy of sleep apnea event detection and OSA diagnosis.
• The proposed method is particularly effective in characterizing non-desaturating sleep apnea events.
• This approach offers a promising tool for more comprehensive and accurate sleep apnea assessment in clinical practice.