Identification of Atrial Fibrillation Biomarkers through Virtual Populations with Anatomical and Electrophysiological Variability

Giada Sira Romitti1, Marí­a Termenón Rivas1, Alejandro Liberos2, Miguel Rodrigo3
1CoMMLab, Universitat de València, 2Universitat de València, 3Universitat de València


Abstract

Atrial fibrillation (AF) is the most common sustained arrhythmia, and its growing prevalence underscores the need for personalized treatment strate-gies. While virtual cohorts offer a promising solution, many studies focus on anatomical or electrophysiological variability in isolation. We evaluated AF vulnerability in 20 bi-atrial anatomical models using si-nus rhythm and 800 simulations, varying pacing sites, {50, 75, 100, 125}% electrical remodeling, and {25, 50}% structural remodeling. Electrical remodeling significantly increased arrhythmia inducibility, with advanced remodeling yielding a higher incidence (50 ± 19%) than less remodeled substrates (37 ± 19%, p = 0.02). Likewise, reduced diffusion, reflecting impaired tissue conductivity, markedly elevated vulnerability (54 ± 14% vs. 37 ± 16%, p < 0.001), suggesting that AF progression - via ionic remodeling or conduction impairment - leads to more arrhythmogenic substrates, independent of anatomy. Pacing site had limited influence on the relative effects of remodeling and diffusion, although LA stimuli consistently produced higher overall AF inducibility. Anatomical features linked to increased susceptibility included greater lateral atrial extent, longer Bachmann's bundle, and prolonged total activation time (TACT), each associated with up to a 20% rise in AF inducibility. Notably, atria with larger dimensions and higher TACT showed a 22% higher arrhythmia risk, highlighting the interaction between structure and function. These findings emphasize the importance of integrating anatomical and electrophysiological variability in AF modeling to identify predictive biomarkers and support digital twin–based therapies.