Non-intrusive PODI-ROM for patient-specific aortic blood flow in presence of a LVAD device

TL;DR

This paper develops a non-intrusive reduced order model using PODI to simulate patient-specific aortic blood flow affected by LVAD devices, enabling efficient analysis of flow modifications with personalized geometry and boundary conditions.

Contribution

It introduces a patient-specific, non-intrusive ROM framework for simulating LVAD-influenced aortic flow, combining CFD, Windkessel models, and PODI for personalized and efficient analysis.

Findings

ROM accurately replicates full order model results

Framework effectively personalizes simulations from medical imaging

Parametric study reveals flow rate impacts on blood flow patterns

Abstract

Left ventricular assist devices (LVADs) are used to provide haemodynamic support to patients with critical cardiac failure. Severe complications can occur because of the modifications of the blood flow in the aortic region. In this work, the effect of a continuous flow LVAD device on the aortic flow is investigated by means of a non-intrusive reduced order model (ROM) built using the proper orthogonal decomposition with interpolation (PODI) method. The full order model (FOM) is represented by the incompressible Navier-Stokes equations discretized by using a Finite Volume (FV) technique, coupled with three-element Windkessel models to enforce outlet boundary conditions in a multi-scale approach. A patient-specific framework is proposed: a personalized geometry reconstructed from Computed Tomography (CT) images is used and the individualization of the coefficients of the three-element Windkessel models is based on experimental data provided by the Right Heart Catheterization (RCH) and Echocardiography (ECHO) tests. Pre-surgery configuration is also considered at FOM level in order to further validate the model. A parametric study with respect to the LVAD flow rate is considered. The accuracy of the reduced order model is assessed against results obtained with the full order model.