Abstract
Introduction/Objective: To date, polysomnography (PSG) remains the 'gold-standard' diagnosis for sleep disorders. Although, the high cost and intrusive nature of the test limits its reach. Prior research indicates that Instantaneous Heart Rate (IHR) might be an accurate and accessible physiological proxy for sleep measurement (Sridhar et al., 2020, npj Digital Medicine). We seek to validate the sleep staging performance of Deep Learning models (Z3Score®-HRV, Neurobit Technologies Pte Ltd) with the CINC open dataset (N=994 Subjects, Goldberger et al., 2020) using IHR derived from a single-channel ECG.
Methods: Single channel ECG was processed on the cloud using the Z3Score® System (https://www.z3score.com/) and was scored in 30-second epochs.
Results: We achieved an accuracy of 72.8% and a Cohen’s kappa of 0.54 on a 4-class staging (Figure 1A). On sleep vs wake classification, we achieved an overall accuracy of 87.7% with sensitivity of 63.3% and specificity of 93.0% (Figure 1B).
Conclusion: Our results demonstrated the utility of IHR derived from ECG for accurate sleep measurement. Sleep-wake detection was also significantly better than any pure actigraphy-based technologies at present. Our results demonstrate that popular ECG-patches (e.g. Movesense, Suunto, Finland) can be a more convenient and affordable alternative to PSG for sleep measurements.
Footnotes
Cite this article as ERJ Open Research 2021; 7: Suppl. 7, 63.
This is an ERS Lung Science Conference abstract. No full-text version is available. Further material to accompany this abstract may be available at www.ers-education.org (ERS member access only).
- Copyright ©the authors 2021
