Effect of Scar Interpolation Methods on Simulated Ventricular Tachycardia in Infarcted Heart Models

Patient-specific computational cardiac models show promise in identifying ablation targets for myocardial infarction (MI) related ventricular tachycardia (VT).However, these models typically rely on low interslice resolution late gadolinium enhanced cardiac MRI (LGE-CMR), necessitating accurate three-dimensional (3D) interpolation of infarct geometry.We proposed a novel methodology using anisotropic Eikonal diffusion of low-resolution scar along apex-base orientations to reconstruct 3D infarct geometry.Moreover, we extensively evaluated its efficacy against high-resolution manual scar reconstructions and the current baseline scar interpolation method, using metrics of both geometrical and electrophysiological similarity.In a porcine cohort of 7 infarcted left ventricle (LV) models, our method demonstrated superior geometrical accuracy compared to the current baseline, achieving an average Dice similarity coefficient (DSC) of 71.5% 7.0%.However, outcomes of electrophysiology simulations revealed greater deviations in VT induction using the proposed methodology compared to gold-standard scar.This study highlights the importance of precise infarct representation in patient-specific models to enhance their functional accuracy for predicting VT inducibility.