Probing Atrial Substrate With the Triple Extrastimulus Protocol: Impact of Variability on Electrophysiological...

Pulmonary vein (PV) isolation is often insufficient to prevent recurrence in persistent atrial fibrillation, highlighting the need to characterize non-PV substrates sustaining reentry. Hidden slow-regions, unmasked under premature stimulation, have emerged as potential arrhythmogenic markers. This study integrates high-density clinical mapping with structurally personalized left-atrial (LA) simulations to assess whether in silico models can reproduce conduction changes elicited by a triple-extrastimulus (TE) protocol. Ten anatomically detailed LA models were built from CT imaging and CARTO mapping, incorporating fibrosis distributions derived from sinus-rhythm bipolar voltage. To account for modelling uncertainties, fibrosis density and global conductivity were systematically varied, generating ten functional variants per patient. Clinical and simulated activation patterns were compared evaluating local activation times (LAT), bipolar amplitude (V_max), and electrogram fractionation. Simulations reproduced key conduction features, achieving moderate LAT agreement (correlation coefficient up to 0.75), with low-to-mild fibrosis improving correspondence. However, variations in fibrosis density produced modest changes, indicating limited sensitivity of global activation patterns. In contrast, V_max and fractionation showed limited agreement (correlation coefficients $< $ 0.4 and 0.25, respectively), reflecting sensitivity to microstructural detail and spatial uncertainty. Notably, the models did not consistently reproduce TE-induced conduction changes observed clinically, underscoring the need for improved restitution modelling to better capture beat-dependent conduction dynamics. Overall, personalized LA models captured global conduction behaviour, but were weakly constrained by clinical measurements, highlighting uncertainty associated with computational models construction. These results support personalization strategies grounded in functional markers, such as restitution or frequency-domain features, combined with coarse structural information to better identify arrhythmogenic substrates.