Estimation of atrial fibrillatory waves from one-lead ECGs using principal component analysis concepts

Abstract

In this contribution is presented a new method for estimating the atrial activity (AA) from one-lead atrial fibrillation (AF) ECGs. This methodology is appropriate for holter signals, where the reduced number of leads is insufficient to exploit the spatial information of the ECG. The proposed approach is based on principal component analysis (PCA) concepts, taking advantage from the time dependence of ventricular and atrial components. The principal components are extracted from the analysis of the ECG at successive cardiac beats, obtaining ventricular and nonventricular related components. The AA is then reconstructed by mapping back all nonventricular components at each cardiac beat. This methodology has been tested and validated using a significative database with simulated and real AF recordings. Its main advantage respect to adaptive template subtraction is its robustness to variations in the QRST shape, thus minimizing the QRST residual in the estimated AA.