Atrial fibrillation detection by multi-lead ECG processing at the edge

Atrial fibrillation is one of the most common arrhythmia events potentially causing heart failures and thrombosis. Recently, many healthcare applications have been developed with the aim to provide a reliable real-time detection of such abnormal heartbeat behavior. The largest part of current solutions considers signal processing applied to electrocardiographic (ECG) segments recorded with wearable devices/sensors and specifically tailored to the number of available ECG leads. Differently, in this contribution, we present a lightweight machine learning algorithm for the analysis of ECG signals and atrial fibrillation detection, easily adaptable for both single and multi-lead architectures. Furthermore, we describe how the proposed scheme can be implemented with an edge computing approach that paves the way toward smart healthcare at home and remotely in general. In such framework, signal processing is not necessarily performed on a unique device, but it is conveniently split so as to let fast operations at the network edge, with storage and heavy computing being instead handled at cloud server side.