Novel Event-Based Peripheral Oxygen Saturation Metrics Provide Complementary Information for OSA Classification.

Area of Science:

• Sleep Medicine
• Biomedical Engineering
• Cardiorespiratory Physiology

Background:

• The Apnea-Hypopnea Index (AHI) is the standard for Obstructive Sleep Apnea (OSA) severity but lacks detailed oxygen desaturation information.
• Current metrics do not capture the dynamics of oxygen level changes during apneic events.

Purpose of the Study:

• To develop and validate novel event-based features quantifying oxygen desaturation and reoxygenation dynamics.
• To integrate these features with existing metrics for improved OSA stratification.

Main Methods:

• A geometric modeling framework was used to extract Event Falling Rate (EFR), Event Recovery Rate (ERR), and Event Baseline Shift (EBS).
• 800 polysomnography recordings from the Sleep Heart Health Study were analyzed.
• Machine learning classifiers integrated these features with established oximetry indices.

Main Results:

• Event-based features showed significant correlations with OSA severity (p < 0.001).
• An integrated model achieved high classification accuracies: 88% (binary), 76% (three-class), and 72% (four-class).
• EFR and ERR were identified as key features for OSA stratification, outperforming traditional metrics.

Conclusions:

• Novel event-based features provide crucial information on hypoxic transition rates and stability, complementing the AHI.
• This framework enables enhanced, personalized OSA assessment using interpretable features from pulse oximetry data.