Discrimination Between CFAEs of Paroxysmal and Persistent Atrial Fibrillation With Simple Classification Models of Reduced Features

Abstract

[EN] A significant number of variables to discriminate between paroxysmal and persistent atrial fibrillation (ParAF vs. PerAF) has been widely exploited, mostly assessed with statistical tests aimed to suggest adequate approaches for catheter ablation (CA) of AF. However, in practice, it would be desirable to utilize simple classification models readily understandable. In this work dominant frequency (DF), AF cycle length (AFCL), sample entropy (SE) and determinism (DET) of recurrent quantification analysis were applied to recordings of complex fractionated atrial electrograms (CFAEs) of AF patients, aimed to create simple models to discriminate between ParAF and PerAF. Correlation matrix filters removed redundant information and Random Forests ranked the variables by relevance. Next, coarse tree models were built, optimally combining high-ranking indexes, and tested with leave-one-out cross-validation. The best classification performance combined SE and DF with an Accuracy (Acc) of 88.2% to discriminate ParAF from PerAF, while the highest single Acc was provided by DET reaching 82.4%. Hence, careful selection of reduced sets of features feeding simple classification models is able to discriminate accurately between CFAEs of ParAF and PerAF