Abstract

Quantifying the biomechanical behavior of cardiovascular tissues holds immense potential to enhance our understanding of (i) smooth and cardiac muscle cell responses to mechanical cues and (ii) how structural and compositional alterations within these tissues impact their overall functional behavior. Such insights are crucial to unravel disease mechanisms (e.g. atherosclerosis, aneurysm, myocardial infarction), to optimize and personalize current treatment strategies (e.g. annuloplasty ring or stent sizing), and to develop novel medical devices (e.g. artificial heart valves or vascular grafts). Such studies depend critically on the accuracy of the underlying constitutive models, which govern the thermodynamically consistent relationship between the tissue’s deformation and internal stress state. Given cardiovascular tissue’s highly non-linear, transverse isotropic or even orthotropic mechanical behavior and a vast ever-increasing library of potential constitutive models, identifying the most accurate material model can be a challenging procedure prone to significant user bias. To resolve this subjectivity and democratize computational engineering analysis for all, we leverage constitutive artificial neural networks and machine learning to automate and democratize constitutive model discovery for these intricate materials. Based on arterial and myocardial biaxial tensile and triaxial shear testing data, our framework autonomously identifies the optimal material models and parameters from a library of over 232 = 4,294,967,296 possible material models. By seamlessly integrating the discovered material models into a universal material subroutine for (in)compressible (an)isotropic tissues, our work advances the implementation of these models into cardiovascular finite element simulations. This not only enhances user-friendliness and robustness but also mitigates the vulnerability to human error. The automated approach signifies a significant stride towards a more inclusive, user-friendly, and accurate framework for cardiovascular biomechanics simulations, ultimately contributing to improved medical treatments for cardiovascular diseases.