In-silico insights into the role of patient-specific fibrosis in personalised therapy selection for atrial fibrillation

Abstract Introduction Current treatments for Atrial Fibrillation (AF) follow a standardised approach that overlooks inter-patient variability in fibrosis and its effect on therapeutic outcomes. A computational framework integrating patient-specific anatomical, fibrotic, and electrophysiological data is needed to assess the effect of fibrosis on therapeutic outcomes and personalise treatment strategies. Purpose To assess the effect of patient-specific fibrosis on the success rates of anti-arrhythmic drugs (AAD) and pulmonary vein isolation (PVI) using personalised biatrial models derived from imaging data. Methods Patient-specific biatrial models (n=69) were created from late gadolinium-enhanced MRI scans using manual segmentation (Fig.1a). Pre-processing included fibrosis quantification, landmark selection and integration of atrial fibre data from a fibre atlas (Fig.1b). Fibrotic masks were generated using normalised blood pool intensities from LGE-MRI images. PVI was simulated by creating circumferential electrically inert lesions around the pulmonary veins (PV) (Fig.1c). Three AAD (vernakalant, amiodarone acute, chronic, and flecainide) were virtually administered via ionic modulation in CARPentry software. Finite element simulations were performed to compare the effects of AAD, PVI, and their combination on AF termination in fibrotic and non-fibrotic anatomies (Fig.2a). Phase singularity (PS) maps were generated to identify re-entry sites and AF wavefront break-up under different treatments. Results Fibrotic anatomies showed significantly higher success rates in AF termination compared to non-fibrotic anatomies across all therapies (p≤0.05) (Fig.2b). The success rates of AAD in fibrotic anatomies were in strong agreement with the ESC 2024 guideline acute success rates (Fig.2b): flecainide showed the highest success rate (58%, ESC 50%-60%), followed by vernakalant (51%, ESC 50%) and amiodarone (30% acute, 39% chronic; ESC 44%). The combination of PVI and AAD did not significantly improve success rates compared to AAD alone in fibrotic anatomies, although this does not incorporate the effects of PV ectopic beat elimination on AF inducibility. Fibrotic anatomies had significantly higher baseline PS counts (p≤0.001), indicating that fibrosis contributes to a more unstable AF substrate (Fig.2a). A negative correlation was found between right atrial (RA) surface area and AAD success rate (rho=-0.42, p≤0.05), with larger RA surface areas associated with reduced AAD effectiveness. Increased PS localisation in the RA with AAD treatments suggests that RA surface area could serve as a potential biomarker for predicting treatment outcomes. Conclusions This work highlights the added value of fibrosis-driven personalisation in therapeutic decision-making. It shows how incorporating patient-specific fibrosis more closely aligns success rates with clinical guidelines, offering a framework for personalised treatment.Biatrial model generation Effect of personalisation