Prediction of Atrial Fibrillation termination by approximate entropy in the time-frequency domain

Given the high prevalence of Atrial Fibrillation (AF) among adult population, to distinguish between these AF episodes that terminate spontaneously and those that persist if no external action is carried out becomes a subject of great clinical interest. In this matter, the complexity analysis of mathematical sequences of parameters obtained from time-frequency distributions of electrocardiograms are useful to recognize the type of AF. The parameters which complexity has been analyzed are the mean peak frequency of atrial activity and the spectral concentration. The complexity measurements of the sequences have been made using the Approximate Entropy, repeating the same procedure for nine distinct time-frequency distributions. The possibility of classifying correctly the AF episodes depends directly on the selected time-frequency distribution.