Abstract
Cardiomyocyte growth can occur in both physiological (exercised-induced) and pathological (e.g., volume overload and pressure overload) conditions leading to left ventricular (LV) hypertrophy. Studies using animal models and histology have demonstrated the growth and remodeling process at the organ level and tissue-cellular level, respectively. However, the driving factors of growth and the mechanistic link between organ, tissue, and cellular growth remains poorly understood. Computational models have the potential to bridge this gap by using constitutive models that describe the growth and remodeling process of the myocardium coupled with finite element (FE) analysis to model the biomechanics of the heart at the organ level. Using subject-specific imaging data of the LV geometry at two different time points, an FE model can be created with the inverse method to characterize the growth parameters of each subject. In this study, we developed a framework that takes in vivo cardiac magnetic resonance (CMR) imaging data of exercised porcine model and uses FE and Bayesian optimization to characterize myocardium growth in the transverse and longitudinal directions. The efficacy of this framework was demonstrated by successfully predicting growth parameters of 18 synthetic LV targeted masks which were generated from three LV porcine geometries. The framework was further used to characterize growth parameters in 4 swine subjects that had been exercised. The study suggested that exercise-induced growth in swine is prone to longitudinal cardiomyocyte growth (58.0 ± 19.6% after 6 weeks and 79.3 ± 15.6% after 12 weeks) compared to transverse growth (4.0 ± 8.0% after 6 weeks and 7.8 ± 9.4% after 12 weeks). This framework can be used to characterize myocardial growth in different phenotypes of LV hypertrophy and can be incorporated with other growth constitutive models to study different hypothetical growth mechanisms.