Linear and non-linear parameters for the classification of atrial fibrillation episodes from intra-atrial signals

In this study linear and non-linear indexes for the characterization of the dynamics in both intra-atrial (IA) signals and atrial period (AP) series are presented and used to discriminate among different atrial fibrillation (AF) episodes. The level of predictability (LP) of AP series and its regularity (R) were computed through the identification of an autoregressive model of the AP series and through the estimation of corrected conditional entropy (CCE), respectively. In addition non-linear synchronization (NLS) patterns were analyzed through indexes based on the mutual corrected conditional entropy (MCCE). Results evidence the capability of the proposed parameters to detect and quantify the different dynamics of IA signals during different type of AF episodes.