Integrated multi-modal data analysis for computational modeling of healthy and location-dependent myocardial infarction conditions in porcine hearts

Abstract

Porcine hearts are widely used for preclinical cardiac evaluation. Computational models, by effectively integrating comprehensive experimental data, often reinforce this preclinical assessment. Using extensive multi-modal data, we developed swine ventricular digital twins for healthy and chronic myocardial infarction (MI) conditions to investigate the roles of the cardiac conduction system (CS), spatial repolarization heterogeneities, cardiomyocyte orientation, cell-to-cell coupling, and MI characteristics on ventricular function. We analyzed cardiac magnetic resonance (CMR) images, electrocardiograms (ECGs), and optical (OM) and electroanatomical mapping from 5 healthy and 10 MI pigs. CS architectures were built from OM and ECG recordings. Myocardial fiber orientation, action potential characteristics, and cell-to-cell conductivity in MI tissue were defined from OM and CMR data. Simulated ECGs for healthy and MI models of left anterior descending and left circumflex occlusions were compared to experimental ECGs and used to assess MI-induced changes. Personalized fiber orientation minimally affected electrophysiology, with conduction velocity (CV) and action potential duration (APD) changing less than 3.6% with respect to standard orientation. Accurate CS and repolarization heterogeneities reproduced depolarization (0.76 QRS similarity) and repolarization (0.74 T-wave similarity) patterns. Incorporating experimentally guided MI-induced alterations enabled replication of MI depolarization and repolarization features (relative errors: 0.5% CV, 2.9% APD), yielded realistic T-wave morphologies (0.63 similarity), and revealed ECG patterns specific to vessel-dependent occlusions. Thus, by integrating extensive multi-modal data, we advance porcine cardiac digital twins and demonstrate the influence of key structural and electrophysiological parameters on healthy and MI heart function, providing a robust computational framework for mechanistic and translational applications. Author summaryPigs are commonly used as preclinical models for cardiac evaluation due to their close resemblance to the human heart. Computational cardiac electrophysiology often supports and extends this preclinical assessment. However, the reliability of these in silico representations depends on the effective integration of comprehensive experimental data. Using extensive multi-modal data, we developed swine ventricular digital twins under healthy and chronic myocardial infarction conditions. These models allowed us to investigate the influence of the cardiac conduction system, spatial repolarization heterogeneities, cardiomyocyte orientation, cell-to-cell coupling, and vessel-specific infarction characteristics on ventricular function. Our results show that cardiomyocyte orientation exerts only a minor effect on electrophysiology, with average conduction velocity showing a slight decrease and action potential duration remaining unchanged when comparing standard versus personalized orientations. Accurate representation of conduction system architecture and repolarization heterogeneities enabled closed reproduction of experimental depolarization and repolarization patterns. Furthermore, by integrating experimentally guided reductions in cell-to-cell coupling and inward rectifier potassium current, along with personalized post-infarction activation alterations and a novel porcine cellular model, we were able to faithfully replicate infarction-specific depolarization-repolarization features. These refinements produced realistic T-wave morphologies and revealed electrocardiographic signatures associated with vessel-dependent infarctions, underscoring the translational potential of our approach.