TY - JOUR T1 - Automated sleep apnea detection from instantaneous heart rate using deep learning models JF - ERJ Open Research JO - erjor DO - 10.1183/23120541.sleepandbreathing-2021.78 VL - 7 IS - suppl 7 SP - 78 AU - Z Siting AU - Y Jie Chen AU - K Kishan AU - A Patanaik Y1 - 2021/04/16 UR - http://openres.ersjournals.com/content/7/suppl_7/78.abstract N2 - Introduction/Objective: Sleep apnea is one of the most common sleep-related breathing disorders that affects nearly a billion people across the world (Benjafield et al., 2019, The Lancet). Polysomnography (PSG), the gold standard test for apnea, is expensive and inconvenient. Thus, there is a strong impetus for alternate methods of diagnosis. We seek to validate the performance of a deep learning based apnea detection system using instantaneous heart rate (IHR) derived from single-channel electrocardiogram (ECG) (Z3ScoreĀ®-HRV, Neurobit Technologies Pte Ltd) on the CINC open dataset (Goldberger et al., 2020).Methods: Single channel ECG was processed on the cloud using the Z3ScoreĀ® System (https://www.z3score.com/). Apnea-Hypopnea Index (AHI) estimation was based on total sleep-time derived from the IHR.Results: Estimated AHI from Z3score correlated very strongly (r=0.85) with true AHI from PSG (Figure 1A). Sensitivity, specificity and AUC of binary classification at various diagnostically relevant AHI thresholds are also presented in figure 1B.Conclusion: Our results demonstrated the feasibility of apnea event detection using only IHR derived from ECG channel. This allows ECG-patch based devices (e.g. Z3Pulse,Neurobit Technologies Pte Ltd) to be used as a more convenient and affordable alternative for Sleep apnea screening.FootnotesCite this article as ERJ Open Research 2021; 7: Suppl. 7, 78.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). ER -