Minimizing discordances in automated classification of fractionated electrograms in human persistent atrial fibrillation

Abstract

Ablation of persistent atrial fibrillation (persAF) targeting complex fractionated atrial electrograms (CFAEs) detected by automated algorithms has produced conflicting outcomes in previous electrophysiological studies. We hypothesize that the differences in these algorithms could lead to discordant CFAE classifications by the available mapping systems, giving rise to potential disparities in CFAE-guided ablation. This study reports the results of a head-to-head comparison of CFAE detection performed by NavX (St. Jude Medical) versus CARTO (Biosense Webster) on the same bipolar electrogram data (797 electrograms) from 18 persAF patients. We propose revised thresholds for both primary and complementary indices to minimize the differences in CFAE classification performed by either system. Using the default thresholds [NavX: CFE-Mean ≤ 120 ms; CARTO: ICL ≥ 7], NavX classified 70 % of the electrograms as CFAEs, while CARTO detected 36 % (Cohen’s kappa κ ≈ 0.3, P < 0.0001). Using revised thresholds found using receiver operating characteristic curves [NavX: CFE-Mean ≤ 84 ms, CFE-SD ≤ 47 ms; CARTO: ICL ≥ 4, ACI ≤ 82 ms, SCI ≤ 58 ms], NavX classified 45 %, while CARTO detected 42 % (κ ≈ 0.5, P < 0.0001). Our results show that CFAE target identification is dependent on the system and thresholds used by the electrophysiological study. The thresholds found in this work counterbalance the differences in automated CFAE classification performed by each system. This could facilitate comparisons of CFAE ablation outcomes guided by either NavX or CARTO in future works.Electronic supplementary materialThe online version of this article (doi:10.1007/s11517-016-1456-2) contains supplementary material, which is available to authorized users.

An interactive platform to guide catheter ablation in human persistent atrial fibrillation using dominant frequency, organization and phase mapping

“…Additionally, other tools can be integrated in the USURP-GUI to allow more comprehensive studies of multiple aspects of the arrhythmia, such as HDF histogram mapping to investigate the DF spatiotemporal behaviors [53], recurrent patterns of HDF algorithms [54] and up-to-date classification techniques for CFAE [55]. The preliminary outcome of using this platform was reported [56].…”

Contributing Factors Concerning Inconsistencies in Persistent Atrial Fibrillation Ablation Outcomes

“…The two commercial systems most frequently used in clinical practice are the NavX and the CARTO (Biosense Webster, Diamond Bar, California) [4]. Those systems provide primary indices to assess CFAE objectively [NavX: CFE-Mean; CARTO: Interval Confidence Level (ICL)], and complementary indices to further support the electrophysiology procedure [NavX: CFE-StdDev; CARTO: Average Complex Interval (ACI), Shortest Complex Interval (SCI)].…”

“…To compare both systems, the 797 AEGs with their respective CFE-Mean and CFE-StdDev were exported from NavX. The ICL, ACI and SCI, as defined by CARTO, were calculated offline with a validated (100% agreement) MATLAB algorithm [4]. CFAE classification was performed by NavX and CARTO using their default clinical thresholds (CFEMean ≤ 120 ms; ICL ≥ 7) [8].…”

“…Nine descriptors were measured from the 207 pairs of AEGs, accordingly: CFE-Mean, CFEStdDev, ICL, ACI, SCI [4], sample entropy (SampEn) [10], peak-to-peak (PP) [11], dominant frequency (DF) [12] and organization index (OI) [13]. Multivariate analysis of variance (MANOVA) and linear discriminant analysis (LDA) were used to test the differences between the AEGs before and after PVI+RL using all descriptors.…”

“…Different companies have developed algorithms based on different features of the atrial electrogram (AEG). Inconsistencies between these algorithms could lead to discordant CFAE classifications by the available systems [4]. Additionally, while studies support that some CFAEs truly represent local AF drivers, others suggest CFAEs are resultant from distant AF drivers [5,6].…”