Unifying automated fractionated atrial electrogram classification using electroanatomical mapping systems in persistent atrial fibrillation studies

Ablation targeting complex fractionated atrial electrograms (CFAE) for treating persistent atrial fibrillation (persAF) has shown conflicting results. Differences in automated algorithms embedded in NavX (St Jude Medical) and CARTO (Biosense Webster) could influence CFAE target identification for ablation, potentially affecting ablation outcomes. To evaluate this effect, automated CFAE classification performed by NavX and CARTO on the same bipolar electrograms from 18 persAF patients undergoing ablation was compared. Using the default thresholds, NavX classified 69±5% of the electrograms as CFAEs, while CARTO detected 35±5%% (Cohen's kappa κ≈0.3, P<0.0001). Both primary and complementary metrics for each system were optimized to balance CFAE detection for both systems. Using revised thresholds found from receiver operating characteristic curves, NavX classified 45±4%, while CARTO detected 42±5% (κ≈0.5, P<0.0001). Our work takes a first step towards the optimization of CFAE detection between NavX and CARTO by providing revised thresholds to reduce differences in CFAE classification. This would facilitate direct comparisons of persAF CFAE-guided ablation outcome guided by NavX or CARTO.