Modeling of Atrial Electrophysiology

Our methodology for computational modeling at the cell and tissue scale of the fibrotic and nonfibrotic atrial electrophysiology can be found in previously published papers. ^{7 , 22 , 23} Briefly, a human atrial action potential model ²⁴ was used to represent membrane kinetics in nonfibrotic myocardium, with the addition of parameter modifications to fit clinical monophasic action potential recordings from patients with AF ^{7 , 25} (I _Kur, Ito, I _CaL decreased by 50%, 50%, and 70%, respectively). In fibrotic regions, further action potential modifications were implemented (I _CaL, I _Na, and I _K1 decreased by 50%, 40%, and 50%, respectively) to represent elevated transforming growth factor‐β, ²⁶ as in prior studies. ^{7 , 11 , 20} At the tissue scale, we applied conductivity tensor values of σ _L=0.409 S m⁻¹ the longitudinal direction and σ _T=0.0820 S m⁻¹ in the transverse, as in our previous study. ²⁰ These conductivity tensor values correspond to physiologically realistic conduction velocity values of 71.49 and 37.14 cm s⁻¹ (longitudinal and transverse). ^{27 , 28 , 29} In fibrotic tissue, the overall conductivity was reduced, and the anisotropy ratio was exaggerated from 5:1 to 8:1 (σ _L = 0.177 S m⁻¹; σ _T=0.0221 S m⁻¹) to represent the effects of interstitial fibrosis and gap junction remodeling. ^{7 , 30 , 31} Prior work has suggested that conduction velocity, fibrosis representation, and action potential duration can markedly affect driver localization dynamics. ^{32 , 33 , 34} Thus, to ensure the set of potential driver sites produced by our analysis was fairly comprehensive in this manner, we repeated all simulations in postablation models with 4 additional conductivity sets (see Figure S1), chosen to increase/decrease conduction velocity by ±10% or ±20%.