Low Complexity Elasticity Models for Cardiac Digital Twins

Abstract: This article introduces CHESRA (Cardiac Hyperelastic Evolutionary Symbolic Regression Algorithm), a novel data-driven framework for automatically designing low-complexity hyperelastic strain energy functions (SEF) for cardiac digital twins. Cardiac digital twins require accurate yet personalized models of heart tissue elasticity, but existing SEF often suffer from high parameter variance, hindering personalization. CHESRA addresses this by using an evolutionary algorithm to derive SEF that balance accuracy and simplicity, while using a normalizing loss function to enable learning from multiple experimental datasets. CHESRA identified two novel SEF, named psi_CH1 and psi_CH2, which use only three and four parameters, respectively, and achieve high accuracy in fitting experimental data. Results show that the parameters of psi_CH1 and psi_CH2 can be estimated more consistently than those of the state-of-the-art SEF when using tissue data, with similar improvements observed for psi_CH1 in a three-dimensional digital twin. CHESRA's utility for generating simple, generalizable SEF makes it a promising tool for advancing cardiac digital twins and clinical decision-making.