Abstract
Introduction
There is a substantial need for an accurate and easy-to-use tool for obstructive sleep apnea (OSA) assessment. Belun Ring Platform (BRP), a novel photoplethysmography (PPG)-based home sleep apnea testing system with a proprietary deep learning algorithm, has been shown to have good sensitivity and specificity in predicting OSA in subjects without significant comorbidities and medications known to affect heart rate (HR). In this study, we further tested its performance in subjects referred for in-lab polysomnography (PSG) assessment of sleep disorders without excluding those with non-arrhythmia comorbidities or the subjects on HR-affecting medications.
Methods
PSG was recorded simultaneously with the Ring in the sleep lab and the studies were manually scored by certified sleep technicians according to the AASM Scoring manual version 2.4. Exclusion criteria include age <18, unstable cardiopulmonary status, recent hospitalization within 30 days, significant arrhythmias, baseline HR <50 or >100, home oxygen use, pacemaker/defibrillator, post-cardiac transplantation or Left ventricular assist device.
Results
A cohort of 78 individuals (26 males and 52 females, age 50.5) were studied with 26 taking HR-affecting medications. Of these, 35 (45%) had AHI < 5; 14 (18%) had AHI 5-15; 15 (19%) had AHI 15-30; 14 (18%) had AHI > 30. The Ring-REI correlated well with the PSG-AHI (r =0.83, P <0.001). The accuracy, sensitivity, specificity in categorizing AHI >15 were 0.808, 0.931, and 0.735 respectively. The positive predictive value, negative predictive value, positive likelihood ratio, and negative likelihood ratio were 0.675, 0.947, 3.509, and 0.094 respectively. The use of HR-affecting medications did not significantly affect the sensitivity and specificity of BRP in predicting OSA (P =0.16 and 0.44 respectively).
Conclusion
BRP is promising as a reasonable tool for OSA assessment and can potentially be incorporated into a broad spectrum of clinical practices for identification of patients with OSA.
Support
This study is supported by a Grant from Belun Technology Company Limited.