Introduction of an expert system for the discrimination of local pulmonary vein and atrial far field signals

Abstract

Background: Discrimination of local and far field potentials during sinus rhythm and atrial fibrillation (AF) is essential for successful pulmonary vein (PV) isolation. We sought to introduce an expert system for the classification of electrophysiologic PV signals.

Methods: For the expert system database, we analyzed ablation procedures of 50 patients with paroxysmal and persistent AF. Standard circumferential catheters and bipolar recordings were required. In a prospective trial, the expert system was compared with the performing electrophysiologists' classifications of potentials during 15 procedures. A total of 1,343 recordings of local PV and far field signals were validated by the sudden disappearance of local potentials during ablation, the presence of dissociated PV activity, and pacing maneuvers. A fast Fourier transform was applied to the individual potentials. Analysis continued in the amplitude and phase representation.

Results: Four parameters significant (p < 0.001) for classification were identified and entered a logistic regression model. Overall sensitivity and specificity of the model was 87% with minor, nonsignificant variations for individual PVs and different underlying rhythms. Concordance with ad hoc electrophysiologists' classification of local potentials was 70%, which increased during post hoc analysis to 86% since classification of 14% of the potentials had to be revised. For these potentials, the expert system correctly predicted their local origin in 86%.

Conclusion: An expert system for the evaluation of electrophysiologic signals based on morphology analysis using the Fourier transform is feasible. The ease of use and online availability facilitate a widespread use for AF ablation procedures.