Different Types of Constitutive Parameters Red Blood Cell Membrane Based on Machine Learning and FEM

Abstract

Research on mechanical response of single red blood cells (RBCs) to mechanical stimuli and the complex material properties of erythrocyte membranes is significant. This work proposes a novel procedure that combines nonlinear finite element method and two machine learning algorithms including Two-Way Deepnets and XGboost together with experiments to identify the hyper elastic material parameters of erythrocyte membranes. Finite element models were established to simulate the stretching process of erythrocyte optical tweezers with different constitutive material parameters from three constitutive models. And the results from the finite element analysis were carried out to generate the training sets for the neural networks. In order to validate the predictions in great detail, the finite element response curves based on the three groups of predicted constitutive parameters are compared with the experimental data. The comparison results show that the Two-Way Deepnets model has performed better efficiency and accuracy and that Reduced Polynomial can describe more precisely the hyperelastic properties of the erythrocyte membrane in the range of experimentally obtained characteristics of single RBCs. This research provides new insights into the identification of constitutive parameters of biological cell membranes, which is crucial for the future research on mechanical mechanisms of the biological cells.