Automated prediction of the apnea-hypopnea index using a wireless patch sensor

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Signal Processing

Background:

• Polysomnography (PSG) is the standard for diagnosing sleep apnea syndrome (SAS) by manually calculating the apnea-hypopnea index (AHI).
• Manual AHI quantification is time-consuming and resource-intensive, necessitating more efficient diagnostic approaches.

Purpose of the Study:

• To develop and validate an automated algorithm for estimating AHI using data from a disposable wireless patch sensor.
• To assess the accuracy of the algorithm in classifying sleep apnea severity.

Main Methods:

• A disposable HealthPatch(TM) sensor was used to collect physiological data from 53 volunteers during overnight PSG studies.
• Features derived from heart rate variability, respiratory signals, posture, and movement were analyzed in 150-second epochs.
• A Linear Support Vector Machine classifier and quadratic regression were employed to detect events and estimate AHI values.

Main Results:

• The algorithm achieved high linear correlation coefficients (0.87-0.92) between predicted and reference AHI values across three sensor locations.
• Classification accuracy for distinguishing between mild (AHI<15) and moderate-to-severe (AHI≥15) sleep apnea ranged from 82.9% to 89.4%.

Conclusions:

• Overnight physiological monitoring with a wireless patch sensor provides an accurate estimation of AHI.
• This automated approach offers a promising, non-invasive method for sleep apnea assessment.