Utilizing a Gaussian Process surrogate model to address Fluid–Structure Interaction issues in evaluating arterial wall stiffness associated with measurement data

This paper discusses the creation of a surrogate model using Gaussian Process Regression to emulate the Fluid–Structure Interaction (FSI) model. Developed model was used to determine stiffness of arteries wall. Here both experimental data as well as numerical results were used to develop surrogate approach. This surrogate is employed for global variance-based uncertainty quantification and sensitivity analysis and is integrated into a hierarchical Bayesian model for parameter estimation. This approach allows for complete probabilistic and concurrent inference of both individual and group-level parameters. While FSI models are frequently used in cardiovascular modeling, they are challenged by lengthy computation times. Developing a surrogate model necessitates thousands (or even hundreds of thousands) of model evaluations, posing a significant demand on computational resources or patience for an average FSI model. The methodology presented here effectively reduces computational costs while maintaining expected precision.