Measuring the complexity of atrial fibrillation electrograms.

ABSTRACT

INTRODUCTION: Complex fractionated atrial electrograms (CFAE) have been identified as targets for atrial fibrillation (AF) ablation. Robust automatic algorithms to objectively classify these signals would be useful. The aim of this study was to evaluate Shannon's entropy (ShEn) and the Kolmogorov-Smirnov (K-S) test as a measure of signal complexity and to compare these measures with fractional intervals (FI) in distinguishing CFAE from non-CFAE signals. METHODS AND RESULTS: Electrogram recordings of 5 seconds obtained from multiple atrial sites in 13 patients (11 M, 58 +/- 10 years old) undergoing AF ablation were visually examined by 4 independent reviewers. Electrograms were classified as CFAE if they met Nademanee criteria. Agreement of 3 or more reviewers was considered consensus and the resulting classification was used as the gold standard. A total of 297 recordings were examined. Of these, 107 were consensus CFAE, 111 were non-CFAE, and 79 were equivocal or noninterpretable. FIs less than 120 ms identified CFAEs with sensitivity of 87% and specificity of 79%. ShEn, with optimal parameters using receiver-operator characteristic curves, resulted in a sensitivity of 87% and specificity of 81% in identifying CFAE. The K-S test resulted in an optimal sensitivity of 100% and specificity of 95% in classifying uninterpretable electrogram from all other electrograms. CONCLUSIONS: ShEn showed comparable results to FI in distinguishing CFAE from non-CFAE without requiring user input for threshold levels. Thus, measuring electrogram complexity using ShEn may have utility in objectively and automatically identifying CFAE sites for AF ablation.