BACKGROUND: Ablation outcomes in long-standing persistent atrial fibrillation (AF) remain limited by diffuse substrate. Although AI-guided spatiotemporal dispersion (STD) mapping identifies complex electrograms, specificity declines with increasing AF duration. Peak frequency (PF) mapping, derived from high-resolution near-field signals, may better pinpoint functionally critical sites. OBJECTIVE: To determine whether PF combined with conduction velocity (CV) mapping improves localisation of AF termination sites during pulsed field ablation (PFA). METHODS: In 28 patients with long-standing persistent AF, high-density left atrial mapping (> 30 000 points) was performed using a paddle‑shaped multielectrode mapping catheter and a contemporary electroanatomic mapping system. STD was annotated with VX1 AI; PF and CV maps were generated offline using omnipolar near-field technology. PF, CV, and voltage at termination (T) sites were compared with control sites; ROC analysis identified predictors of termination. RESULTS: AF terminated in 18/28 patients (64%). Termination sites clustered in four regions (anteroseptal LA, anterior LA, inferior LA–CS, RSPV ridge). PF was markedly higher at T sites (389 ± 81 vs. 202 ± 51 Hz; P < 0.001) and showed excellent discrimination (AUC 0.96; ≥340 Hz: sensitivity 88%, specificity 90%), outperforming voltage (AUC 0.59). PF maxima colocalised with CV deceleration zones. At 12 months, arrhythmia recurrence was 18% overall (40% without vs. 5% with acute termination). CONCLUSION: PF mapping during ongoing AF reliably identifies functional substrate and termination sites during PFA, outperforming voltage and complementing AI-STD mapping. Integration of automated PF/CV analysis may enhance outcomes in long-standing persistent AF. Multicentre validation is warranted.
Keywords: Artificial intelligence; Conduction Velocity; Peak Frequency; Persistent atrial fibrillation; Pulsed field ablation; Spatial Temporal Dispersion.