Apneic Event Estimation only using SpO2 Dynamics in Sleep Apnea Patients

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Digital Health

Background:

• Nocturnal pulse oximetry is a potential tool for diagnosing sleep apnea.
• Previous studies established criteria for apnea event detection using blood oxygen saturation (SpO2) changes.
• An apnea-hypopnea index (AHI) estimation model was developed using regression analysis.

Purpose of the Study:

• To verify the performance of a previously developed algorithm for sleep apnea diagnosis.
• To validate the algorithm using data from an external sleep medicine center.
• To assess the potential of the algorithm for mobile healthcare devices.

Main Methods:

• Applied a previously developed algorithm to pulse oximetry data from 15 polysomnographic recordings.
• Extracted quantitative characteristics from SpO2 variations to identify apneic events.
• Estimated the apnea-hypopnea index using regression modeling and compared it to reference values.

Main Results:

• Minute-by-minute apneic segment detection achieved 87.58% average accuracy and a Cohen's kappa of 0.6327.
• The correlation coefficient between the estimated and reference AHI was 0.95, with an average absolute error of 5.02 events/h.
• The algorithm demonstrated meaningful semi-real-time sleep apnea event detection and AHI estimation on external data, albeit with slightly reduced performance.

Conclusions:

• The validated algorithm shows effectiveness in detecting sleep apnea events and estimating AHI from pulse oximetry data.
• The findings support the integration of sleep apnea diagnosis and monitoring functions into wearable mobile healthcare devices.
• This research enhances the user value of mobile health technologies for sleep disorder management.