Intelligent Mobile Electrocardiogram Monitor-empowered Personalized Cardiac Big Data*

Abstract

Smart health big data is quickly driving the healthcare field and bringing numerous new opportunities. Cardiac disease is a leading cause of death worldwide, and the personalized cardiac big data is expected to offer new strategies and possibilities for cardiac health management. The standard 12-lead electrocardiogram (ECG) has been a gold standard of cardiac health measurement for decades. However, there is still lack of effective ways to monitor 12-lead ECG in our daily lives, which is a critical obstacle towards personalized cardiac big data. In this study, we have proposed and validated a mobile 3-lead ECG monitoring system that can reconstruct the standard 12-lead ECG, offering a much greater usability for daily ECG tracking compared with the traditional 12-lead ECG system. Moreover, the system is able to deal with severe motion artifacts during daily physical exercises and yield high-fidelity ECG reconstruction, leveraging a deep recurrent neural network. A multi-stage long short-term memory network has been proposed to reconstruct the robust 12-lead ECG from the noisy 3-lead ECG. This motion artifacts-tolerant ability is highly important, considering that users may perform diverse and random physical activities, which will inevitably contaminate or even corrupt the ECG signal. The reconstruction error is as low as 0.069, and the correlation coefficient is as high as 0.84. This unobtrusive and motion-tolerant mobile ECG monitoring system has been validated on human data and demonstrated the feasibility to continuously establish the personalized cardiac big data. This research is highly encouraging and is expected to be able to significantly advance big data-driven cardiac health management.