Prediction Of Atrial Fibrillation From At-home Single-lead Ecg Signals Without Arrhythmias

This article presents a framework for predicting the occurrence of atrial fibrillation (AF) using a single-lead ECG obtained from a chest patch. The framework combines demographic characteristics, HRV metrics, ectopic beat frequency, and morphologic analysis of the single-lead ECG with a deep learning approach. Results show that the inclusion of deep learning features extracted from the single-lead ECG signal, along with demographic and phenotypic variables, allows for the most accurate prediction of AF within a 14-day period.