Study on Atrial Arrhythmias Optimal Organization Assessment with Generalized Hurst Exponents

Abstract

The aim of the present work is to determine the optimal use of the Generalized Hurst Exponents (H(q)) in the noninvasive organization assessment of atrial arrhythmias. In case of q = 2, H(2) relates to the existence of statistical dependencies in the autocorrelation function of a time series. Therefore, it could be used to estimate the pattern repetitivity of atrial arrhythmias. Given that the most common atrial arrhythmia is atrial fibrillation (AF), this work analyzes optimal computational parameters of H(2) with the aim to predict spontaneous AF termination because the probability of AF termination depends on its organization. For this purpose, a reference database containing non-terminating and terminating AF episodes was analyzed. First, H(2) was computed over all non-overlapping segments ranging from 1 up to 30 seconds in order to determine the optimal data length. Then, since the presence of noise and ventricular residua degrades nonlinear metrics performance, the use of a band-pass filtering stage was evaluated. Finally, only the last seconds of each recording were analyzed to determine the applicability of H(2) to short electrocardiographic (ECG) recordings. After optimal computational parameters selection, H(2) yielded a prediction accuracy of 95 notably better result than previous studies on the same database. Therefore, H(2) can be applied in the study of ambulatory ECG recordings with organization-dependent events.