Efficient Screening in Obstructive Sleep Apnea Using Sequential Machine Learning Models, Questionnaires, and Pulse...

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Artificial Intelligence

Background:

• Obstructive sleep apnea (OSA) is a common sleep disorder impacting breathing.
• Traditional polysomnography is effective but resource-intensive, limiting widespread diagnosis.

Purpose of the Study:

• To develop two sequential machine learning models for efficient OSA screening and severity differentiation.
• To improve diagnostic efficiency compared to traditional methods.

Main Methods:

• Utilized two datasets (SHHS and TVGH) with 8444 and 1229 cases, respectively.
• Developed a Questionnaire Model (demographics, Pittsburgh Sleep Quality Index) and a Saturation Model (blood oxygen saturation parameters).
• Evaluated model performance on an independent test set.

Main Results:

• The Questionnaire Model achieved an F1-score of 0.86.
• The Saturation Model achieved F1-scores of 0.82-0.85 depending on the dataset and parameters used.
• The independent test set showed stable performance, with precision remaining high (0.89) for moderate to severe OSA screening.

Conclusions:

• Sequential machine learning models show promise for at-home OSA screening.
• Optimization of models, particularly identifying key saturation parameters, is crucial for enhanced accuracy.
• These models provide a valuable tool for patients unable to undergo in-laboratory sleep studies.