A Framework for Real-Time Source-Agnostic ECG Monitoring Using Artificial Intelligence

Abstract

Background: Long-term, continuous real-time ECG monitoring for arrhythmia detection has been limited by challenges in development of comfortable, wearable ECG devices, real-time streaming and accurate, automated analysis. Our framework involves three independent but linked methodologies, involving ECG acquisition on a small wearable device, continuous wireless digital transfer of ECG data to a mobile device, and real-time arrhythmia-detection using artificial intelligence (AI). Methods: The current study was performed to compare our automated system against simultaneous ECG monitoring in a coronary care unit, using a standard patient monitoring system. ECGs were de-identified and randomised, and blinded R-R interval measurements were performed using manual digital callipers by three independent cardiologists. Results: ECG acquisition and continuous wireless transmission showed excellent data integrity and no significant dropouts. Storage and encryption of data was robust. The ECG signals acquired by our device appeared visually identical to the standard monitoring system. Agreement of R-R interval measurement between devices and observers was good, with RMSE and Bias at 26 and 6 ms, respectively. The rhythm comparison showed an accuracy of 93%. Conclusion: Our continuous, wearable ECG device with real time wireless data transmission to a mobile device provides robust data integrity, and good agreement compared with a standard ECG monitoring system. The framework provides a suitable platform for automated continuous arrhythmia detection on a smart phone, using AI.