A method to screen obstructive sleep apnea using multi-variable non-intrusive measurements.

Area of Science:

• Sleep Medicine
• Biomedical Engineering
• Acoustics

Background:

• Obstructive sleep apnea (OSA) is a prevalent, underdiagnosed sleep disorder.
• Current diagnosis via polysomnography (PSG) is costly, inconvenient, and unsuitable for community screening.
• Simplified, unattended, and affordable screening methods for OSA are critically needed.

Purpose of the Study:

• To develop and validate a novel, non-intrusive method for screening obstructive sleep apnea (OSA).
• To utilize snoring sounds and neck circumference as primary indicators for OSA detection.
• To provide a cost-effective and accessible screening tool for widespread OSA identification.

Main Methods:

• A multi-feature vector was created, incorporating pitch, formant information, and periodic structure of snore signals, along with neck circumference.
• Snore signals were recorded and analyzed in a sleep laboratory setting.
• A neural network classified OSA/non-OSA status, with validation using K-fold cross-validation and random sub-sampling. A background noise removal technique was also implemented.

Main Results:

• The proposed method achieved a sensitivity of 91 ± 6% and a specificity of 89 ± 5% for detecting OSA (AHI threshold = 15) in a database of 51 subjects.
• The multi-feature vector effectively characterized OSA conditions using snore acoustics and physical measurements.
• The method demonstrated robust performance on test data, indicating its potential for reliable OSA screening.

Conclusions:

• Snoring sound analysis, combined with neck circumference, presents a viable and promising approach for non-intrusive OSA screening.
• This method offers a significant improvement over traditional PSG for community-based screening due to its simplicity and lower cost.
• The developed technique has the potential to increase OSA diagnosis rates by addressing the limitations of current screening methods.