Atrial Fibrillation

Standardizing Single-Frame Phase Singularity Identification Algorithms and Parameters: In the present study, we demonstrate that automated PS detection – and consequently persAF ablation target identification – vary significantly for the same individual, depending on the method being used and parameters being applied. We propose revised parameters that optimize the PS detection performed by the different algorithms according to a clinical “gold standard.” Four algorithms were evaluated – a 2D image node-neighbor; a 3D node-neighbor; a 2D convolutional kernel topological charge; and a 3D topological charge. Optimal parameters were proposed for each algorithm and should be used in future studies to improve the accuracy of PS detection. The 3D topological charge with DBSCAN clustering and proposed parameters has shown the best accuracy. Similarly, the algorithm that estimates topological charge using a convolutional kernel with DBSCAN clustering and proposed parameters should be preferred for uniformed 2D meshes. The present study represents a step toward a unified definition/algorithm of phase-derived PS detection with standardized gradient and spatial thresholds, which is essential to allow objective comparisons of outcomes of rotor ablation for persAF therapy among different research/clinical centers.