Abstract
This document serves as a formal correction to the 2021 study "Detection of obstructive sleep apnea using Belun Sleep Platform wearable with neural network-based algorithm and its combined use with STOP-Bang questionnaire." The correction addresses a data-swapping error in the demographic reporting of the study's cohort, ensuring the integrity of research aimed at making sleep apnea diagnosis more accessible through wearable technology and machine learning.
Introduction
Obstructive Sleep Apnea (OSA) is a prevalent disorder characterized by repeated interruptions in breathing during sleep. Despite its significant health implications, it remains frequently undiagnosed due to the limitations of the current "gold standard" diagnostic tool, Polysomnography (PSG). PSG requires overnight stays in specialized labs with numerous sensors, making it expensive, invasive, and often subject to long waiting lists. The original research sought to validate a more accessible alternative: the Belun Sleep Platform.
Methodology and Technology
The study evaluated a wearable ring-style device that utilizes neural network-based algorithms to identify OSA patterns. Unlike traditional screening methods that often rely on simple threshold triggers (such as a specific dip in oxygen levels), this system uses a neural network to identify complex patterns in physiological data streams that correlate with apnea events.
To enhance diagnostic confidence, the researchers employed a hybrid screening approach, combining objective data from the wearable with subjective clinical data from the STOP-Bang questionnaire, a standard 8-item screening tool used by clinicians. The system's performance was validated against traditional PSG results across various levels of OSA severity (Mild, Moderate, and Severe).
Corrected Demographic Data and Results
The primary purpose of this correction is to rectify the racial distribution data for the study's 78 participants. In the original publication, the counts for "White" and "Asian" participants were inadvertently swapped. Accurate demographic reporting is critical in medical AI to ensure algorithms are developed and tested equitably across different populations.
As shown in the corrected Table 1:
- Cohort Composition: The group consisted of 78 subjects (65% female, 35% male) with a mean age of approximately 51.5 years.
- Racial Distribution: The corrected data shows 30 White participants (38%) and 4 Asian participants (5%), alongside 41 Black participants (53%) and 3 others (4%).
- Clinical Findings: Out of the 78 subjects, 43 (55%) were diagnosed with some form of OSA via PSG. The data highlights a strong correlation between BMI and OSA severity, with the "No OSA" group averaging a BMI of 31.4 compared to 42.5 in the "Mild OSA" group.
Significance of the Findings
The significant p-values (<0.001) for the Apnea-Hypopnea Index (AHI) across severity groups validate that the participants were correctly categorized, supporting the algorithm's underlying training logic. Furthermore, the high statistical significance of STOP-Bang scores across groups suggests that the wearable platform effectively captures the same physiological signals that traditional, more expensive methods do.
Conclusion
While the sample size of 78 subjects suggests a pilot-scale study, the results demonstrate the potential of combining wearable sensors with machine learning to streamline OSA diagnosis. Moving screening from the lab to a natural home environment via wearable rings could significantly improve the detection and management of this common sleep disorder, allowing for longitudinal monitoring of sleep health.
Summary updated