Automatic home OSA screening with pulse oximeters often relies on oxygen saturation (SpO2), which may miss hypopneas associated with arousals. Photoplethysmography (PPG) from pulse oximeters can be processed to extract pulse wave amplitude (PWA) and pulse-to-pulse interval (PPI) which reflect autonomic activation during arousals. However, the role of PWA in automated OSA screening remains unexplored. This study evaluated the added value of PWA for detecting OSA segments, and estimating AHI compared with SpO2 and PPI. From 90 PSG recordings, PWA and PPI were derived from finger PPG. We extracted statistical PWA features to capture amplitude drops, and spectral powers to represent pulse amplitude variability. PPI features included statistics and heart rate variability measures. Support vector machine classifiers with different combinations of PWA, PPI, and SpO2 features were trained to detect 60-second segments with apnea or hypopnea. Performance was assessed at both per-segment and per-subject levels for identifying AHI ≥ 15. Adding PWA to SpO2 improved sensitivity for arousal-related OSA segments from 61.6% to 65.2% and increased per-subject sensitivity from 61.4% to 64.9%. Adding PPI to SpO2 improved per-segment sensitivity for arousal-related OSA segments to 73.3% and increased per-subject sensitivity to 77.2%. Combining PWA with PPI achieved the highest sensitivity for arousal-related segments (77.1%), and when both were added to SpO2, per-subject sensitivity for detecting AHI ≥ 15 reached 80.7%. Both PWA and PPI improved detection of arousal-related segments and contributed to detecting subjects with AHI ≥ 15. However, PPI consistently outperformed PWA. SpO2 remained particularly important for identifying subjects with AHI < 15.
Thuptimdang et al. (Thu,) studied this question.