Wavelet analysis of electrocardiograms to characterize recurrent atrial fibrillation

Abstract

In a substantial number of patients atrial fibrillation (AF) recurs after successful electrical cardioversion. Recurrence of AF is insufficiently predictable by clinical and echocardiographic parameters used in the clinic procedure. In this study some parameters were extracted from the analysis of electrocardiograms (ECGs), in an effort to predict the maintenance of sinus rhythm after cardioversion in patients with persistent AF. The database under study includes some ECG registers undergoing cardioversion with some additional physiological and anatomical information of each patient. After 12 weeks following cardioversion, only 15 (41.6%) of 34 patients maintained sinus rhythm. The ECG recordings were processed using the continuous wavelet transform (CWT) and the discrete wavelet transform (DWT). The method estimated the predictive power with respect to defibrillation outcome of some parameters extracted with CWT and DWT analysis, such as the main frequency peak frequency and the energy of detail coefficients, respectively. A logistic regression model was constructed, determining the calibration with the Homer–Lemeshow test and the discrimination with the area under the receiver operation characteristic (ROC) curve. The proposed methodology demonstrated a diagnostic capability of 82.4% for the prediction of AF recurrence, contributing to an improved interpretation of AF arrhythmias and their relation with recurrence risk.