Signal processing methods for information enhancement in atrial fibrillation : spectral analysis and non-linear parameters

Recently, the research efforts in the context of electrocardiographical recording during atrial fibrillation (AF) has been directed to broaden the understandings on the electrophysiological and structural remodelling occurring during the arrhythmia and on characterizing the different types of AF. Following this line, both surface ECG and endocardial electrograms have been thoroughly studied and a series of linear and non-linear parameters were computed either directly on the electrograms or on the derived activation series. In this paper, we reviewed some signal processing methods used to characterize surface ECG and endocardial electrograms during AF, focusing on spectral and non-linear analysis. In particular, parametric and non-parametric methods for spectral analysis of the residual ECG, i.e. atrial waves obtained from surface ECG after removing ventricular activity, and endocardial recordings are described. The different purposes of spectral analysis (exploring autonomic functions, analysis of spontaneous AF behaviour and predicting therapeutic effects) are illustrated with some examples. In addition, we described some more recent non-linear methods applied to AF, assessing the organization of atrial signals as well as ventricular response in AF. In particular, methods derived from embedding time series and based on entropy computation are illustrated and exemplified.