B-PO03-073 ATRIAL INFORMATION: AUTOMATED ELECTROANATOMIC MAP ANALYSIS FOR PREDICTING ATRIAL FIBRILLATION ABLATION OUTCOMES

Abstract

Outcomes research to identify risk factors for arrhythmia recurrence following AF ablation has historically focused on comorbidities and technologies utilized. The electroanatomic (EA) map generated during the procedure contains innumerable details that are not routinely included in prediction models due to the challenge of manually obtaining and quantifying this data. To develop software that automatically extracts comprehensive metrics from the EA map and synthesizes this with relevant clinical data for machine learning classification algorithms to identify predictors of AF ablation outcome. Innovative software was developed to evaluate EA maps from patients who underwent AF ablation at our tertiary medical center. The software also queried the electronic medical record to retrieve relevant clinical characteristics. We identified 7028 patients who underwent AF ablation from 2010-2020, of which 1779 had a redo procedure. Over 1000 metrics were extracted from available EA maps. Data quality was adjudicated using point density, catheter motion during point collection, and EGM signal:noise ratio. Voltage, dominant frequency, and signal fractionation statistics were computed by wall segment and by dividing the map into equally spaced units, providing insights into atrial heterogeneity. Ablation lesions were assessed by location, stability, duration, force, impedance change, and sequence. Machine learning algorithms are used to classify this data along with clinical characteristics by ablation outcome. Automated EA map analysis adds important procedural metrics to identify predictors of successful AF ablation.