Wavelet sample entropy: A new approach to predict termination of atrial fibrillation

The mechanisms that provoke the eventual termination of some self paroxysmal fibrillation (PAF) episodes still remain unexplained. The aim of this to discriminate between between the groups of terminating (T registers) and non-terminating (N registers) of PAF episodes by using the ECG. A new technique, called wavelet sample entropy (WSE) is proposed. WSE exploits the combination of wavelet transform properties with regularity measure indexes, as sample entropy (SampEn). Results indicae that terminating episodes present lower mean values of SampEn than the non-terminating episodes. These results are consistent with the fact that fibrillatory activity become slower and more organized prior to termination. Both PAF groups were statistically distinguishable, given that the statistic significance obtained by the t-student test equals to 0.001 and 90% of test signals are correctly classified. This new method could improve the efficiency of traditional discrimination techniques, based on time or frequency analysis, and it could be helpful for a better understanding of atrial fibrillation mechanisms.