A model for patient-specific cardiac mechanics simulation is introduced, incorporating a 3-dimensional finite element model of the ventricular part of the heart, which is coupled to a reduced-order 0-dimensional closed-loop vascular system, heart valve, and atrial chamber model. The ventricles are modeled by a nonlinear orthotropic passive material law. The electrical activation is mimicked by a prescribed parameterized active stress acting along a generic muscle fiber orientation. Our activation function is constructed such that the start of ventricular contraction and relaxation as well as the active stress curve's slope are parameterized. The imaging-based patient-specific ventricular model is prestressed to low end-diastolic pressure to account for the imaged, stressed configuration. Visco-elastic Robin boundary conditions are applied to the heart base and the epicardium to account for the embedding surrounding. We treat the 3D solid-0D fluid interaction as a strongly coupled monolithic problem, which is consistently linearized with respect to 3D solid and 0D fluid model variables to allow for a Newton-type solution procedure. The resulting coupled linear system of equations is solved iteratively in every Newton step using 2  ×  2 physics-based block preconditioning. Furthermore, we present novel efficient strategies for calibrating active contractile and vascular resistance parameters to experimental left ventricular pressure and stroke volume data gained in porcine experiments. Two exemplary states of cardiovascular condition are considered, namely, after application of vasodilatory beta blockers (BETA) and after injection of vasoconstrictive phenylephrine (PHEN). The parameter calibration to the specific individual and cardiovascular state at hand is performed using a 2-stage nonlinear multilevel method that uses a low-fidelity heart model to compute a parameter correction for the high-fidelity model optimization problem. We discuss 2 different low-fidelity model choices with respect to their ability to augment the parameter optimization. Because the periodic state conditions on the model (active stress, vascular pressures, and fluxes) are a priori unknown and also dependent on the parameters to be calibrated (and vice versa), we perform parameter calibration and periodic state condition estimation simultaneously. After a couple of heart beats, the calibration algorithm converges to a settled, periodic state because of conservation of blood volume within the closed-loop circulatory system. The proposed model and multilevel calibration method are cost-efficient and allow for an efficient determination of a patient-specific in silico heart model that reproduces physiological observations very well. Such an individual and state accurate model is an important predictive tool in intervention planning, assist device engineering and other medical applications.

For treatment of advanced heart failure, current strategies include cardiac transplantation or blood-contacting pump technology associated with complications, including stroke and bleeding. This study investigated an individualized biventricular epicardial augmentation technology in a drug-induced porcine failing heart model. A total of 11 pigs were used, for the assessment of hemodynamics and cardiac function under various conditions of support pressures and support durations (n = 4), to assess device positioning and function by in vivo computer tomographic imaging (n = 3) and to investigate a minimally invasive implantation on the beating heart (n = 4). Support pressures of 20-80 mm Hg gradually augmented cardiac function parameters in this animal model as indicated by increased left ventricular stroke volume, end-systolic pressures, and decreased end-diastolic pressures. Strong evidence was found regarding the necessity of mechanical synchronization of support end with the isovolumetric relaxation phase of the heart. In addition, the customized, self-expandable implant enabled a marker-guided minimally invasive implantation through a 4 cm skin incision using fluoroscopy. Correct positioning was confirmed in computer tomographic images. Continued long-term survival investigations will deliver preclinical evidence for further development of this concept.

Advances in ventricular assist device (VAD) technology for the treatment of end‐stage congestive heart failure (CHF) are needed to cope with the increasing numbers of patients that cannot be provided with donor hearts for transplantation. We develop and investigate a novel extravascular VAD technology that provides biventricular, epicardial pressure support for the failing heart. This novel VAD concept avoids blood contact that is accompanied with typical complications such as coagulation and infections. To date, in vivo porcine model results with a prototype of the implant exist, further studies to improve the implant's performance and promote its applicability in humans are needed. In this contribution, we present a personalised functional digital twin of the heart, the vascular system, and the novel VAD technology in terms of a calibrated, customized computational model. The calibration procedure is based on patient‐specific measurements and is performed by solving an inverse problem. This in silico model is able to (a) confirm in vivo experimental data, (b) predict healthy and pathologic ventricular function, and (c) assess the beneficial impact of the novel VAD concept to a high level of fidelity. The model shows very good agreement with in vivo data and reliably predicts increases in stroke volume and left ventricular pressure with increasing ventricular support. Furthermore, the digital twin allows insight into quantities that are poorly or not at all amenable in any experimental setup. Conclusively, the model's ability to link integral hemodynamic variables to local tissue mechanical deformation makes it a highly valuable tool for the dimensioning of novel VAD technologies and future treatment strategies in heart failure. The presented in silico twin enhances in vivo studies by facilitating the accessibility and increasing the range of quantities of interest. Because of its flexibility in the assessment of design variants and optimization loops, it may substantially contribute to a reduction of the amount of animal experiments in this and similar settings.

No abstract