Patient-specific input data for predictive modeling of the Fontan operation

Abstract

Personalized blood flow models are used for optimization of the Fontan procedure. In this paper we discuss clinical data for model initialization. Before the Fontan procedure patients undergo CT or MRI examination. Computational domain of interest is reconstructed from this data. CT images are shown to have a better spatial resolution and quality and are more suitable for segmentation. MRI data gives information about blood flow rates  and it is utilized for setting boundary conditions in local 3D hemodynamic models.   We discovered that the MRI data is contradictory and too inaccurate for setting boundary conditions: the error of measured velocities  is comparable with blood velocities in veins. We discuss a multiscale 1D3D circulation model as  potentially suitable for prediction of the Fontan procedure results. Such model may  be initialized with more reliable data (MR measurements of blood flow in aorta and ultrasound examination of easily accessible vessels) and take into account  collateral and fenestration blood flows which are typical for Fontan patients. We have calculated these flow rates for several patients and demonstrated  that such flows occur systematically.