Developing a surrogate model for minimizing the computation expense of pulsating blood flow simulation

Abstract

Three-dimensional blood flow simulations have emerged as a viable option for diagnosing cardiovascular disease, predicting treatment outcomes, and developing and evaluating medical devices. However, since three-dimensional simulations can take a long time to execute, they are limited in clinical practice and optimization and uncertainty quantification methods. To overcome these limitations, various reduced order models are developed using reduced order models, proper orthogonal decomposition methods, machine learning methods and principal component analysis, to list a few. These methods, however, are developed for steady flows and only a few studies have developed reduced order models for pulsatile blood flow simulations. We develop a reduced order model for pulsatile flows using a one-dimensional blood flow assumption and apply to various patient-specific geometries. The reduced order model calculates pressure difference as a function of flow rates but the coefficients are computed based on minimal blood flow simulation results instead of deriving them from geometric relationships. The reduced order model is evaluated for an ideal cylindrical model and extended to patient-specific models constructed from medical image data.