Apnea-hypopnea index estimation using quantitative analysis of sleep macrostructure

Area of Science:

• Sleep Medicine
• Medical Technology
• Data Science

Background:

• Obstructive sleep apnea (OSA) is a common condition linked to sleep macrostructure changes.
• Current applications of this link are limited.
• The apnea-hypopnea index (AHI) is a key metric for OSA assessment.

Purpose of the Study:

• To propose a novel quantitative approach for estimating the AHI using sleep macrostructure.
• To explore new sleep macrostructure parameters beyond conventional ones.

Main Methods:

• Extracted novel sleep macrostructure parameters from 132 polysomnographic recordings.
• Utilized nonlinear regression analysis, focusing on transitions from NREM stage 2 to stage 1 sleep.
• Split data into training and validation sets, with a significant portion having AHI > 5 events/h.

Main Results:

• Nonlinear regression using percentage transition probability from NREM stage 2 to stage 1 was most effective for AHI estimation.
• Achieved a root mean square error of 7.30 events/h in the validation set.
• Demonstrated high diagnostic performance at an AHI cut-off of >= 30 events/h: 90.0% sensitivity, 93.5% specificity, 92.4% accuracy.

Conclusions:

• The developed AHI estimation model shows potential for OSA assessment.
• This method is valuable in situations where respiration signals cannot be obtained or analyzed.
• It enables OSA diagnosis using only sleep macrostructure information.