In silico trials can assess the effectiveness of new therapeutic methods before clinical evaluations. In this study, we propose a computational pipeline to develop personalized digital twins of cardiac ventricles, enabling a robust characterization of ventricular passive behavior with a focus on the right ventricle (RV). Our framework was employed to simulate the hearts of three right ventricular (RV) failure patients. The estimation of biomechanical properties, coupled with finding the unloaded shapes of the ventricles, was performed by minimizing the deviation between the simulated and target RV end-diastolic pressure-volume relationships (EDPVRs). Finite element analysis (FEA) was used to model the ventricles, enabling the virtual addition of an epicardial restraint to the RV. We examined two possible definitions of the target restrained RV EDPVRs: exponential and proportional volume reductions. Bayesian optimization was utilized for both material evaluation and constraint design on a wide range of available biomaterials across varying levels of desired RV volume reduction ratios. Sensitivity analysis indicated that increasing constraint thickness or stiffness shifts the RV EDPVR curve leftward. However, this shifting stops at a specific curve, beyond which further increase in restraint thickness or stiffness cannot reduce RV volumes. A perfect RV EDPVR fit was achieved with the optimal restraint, particularly for the exponentially shifted targets. The optimal thickness increases as the desired reduction ratio rises, which also corresponds to decreased fiber stress and strain in the right ventricular free wall (RVFW). Patient-specific design of epicardial restraints can offer the most effective treatment.
