Yann T. Delorme

A novel multiblock immersed boundary method for large eddy simulation of complex arterial hemodynamics

Computational fluid dynamics (CFD) simulations are becoming a reliable tool to understand hemodynamics, disease progression in pathological blood vessels and to predict medical device performance. Immersed boundary method (IBM) emerged as an attractive methodology because of its ability to efficiently handle complex moving and rotating geometries on structured grids. However, its application to study blood flow in complex, branching, patient-specific anatomies is scarce. This is because of the dominance of grid nodes in the exterior of the fluid domain over the useful grid nodes in the interior, rendering an inevitable memory and computational overhead. In order to alleviate this problem, we propose a novel multiblock based IBM that preserves the simplicity and effectiveness of the IBM on structured Cartesian meshes and enables handling of complex, anatomical geometries at a reduced memory overhead by minimizing the grid nodes in the exterior of the fluid domain. As pathological and medical device hemodynamics often involve complex, unsteady transitional or turbulent flow fields, a scale resolving turbulence model such as large eddy simulation (LES) is used in the present work. The proposed solver (here after referred as WenoHemo), is developed by enhancing an existing in-house high order incompressible flow solver that was previously validated for its numerics and several LES models by Shetty et al. [Journal of Computational Physics 2010; 229 (23), 8802-8822]. In the present work, WenoHemo is systematically validated for additional numerics introduced, such as IBM and the multiblock approach, by simulating laminar flow over a sphere and laminar flow over a backward facing step respectively. Then, we validate the entire solver methodology by simulating laminar and transitional flow in abdominal aortic aneurysm (AAA). Finally, we perform blood flow simulations in the challenging clinically relevant thoracic aortic aneurysm (TAA), to gain insights into the type of fluid flow patterns that exist in pathological blood vessels. Results obtained from the TAA simulations reveal complex vortical and unsteady flow fields that need to be considered in designing and implanting medical devices such as stent grafts.

Summation-by-parts Operators with Minimal Dispersion Error for Accurate and Efficient Flow Calculations

We develop summation-by-parts operators with minimal dispersion errors both near and far from boundaries and interfaces. Such operators are superior to classical stencils for problems involving hig ...

Dynamic Mode Decomposition of Fontan Hemodynamics in an Idealized Total Cavopulmonary Connection

Univentricular heart disease is the leading cause of death from any birth defect in the first year of life. Typically, patients have to undergo three open heart surgical procedures within the first few years of their lives to eventually directly connect the superior and inferior vena cavae to the left and right pulmonary arteries forming the Total Cavopulmonary Connection or TCPC. The end result is a weak circulation where the single working ventricle pumps oxygenated blood to the body and de-oxygenated blood flows passively through the TCPC into the lungs. The fluid dynamics of the TCPC junction involve confined impinging jets resulting in a highly unstable flow, significant mechanical energy dissipation, and undesirable pressure loss. Understanding and predicting such flows is important for improving the surgical procedure and for the design of mechanical cavopulmonary assist devices. In this study, Dynamic Mode Decomposition (DMD) is used to analyze previously obtained Stereoscopic Particle Imaging Velocimetry (SPIV) data and Large Eddy Simulation (LES) results for an idealized TCPC. Analysis of the DMD modes from the SPIV and LES serve to both highlight the unsteady vortical dynamics and the qualitative agreement between measurements and simulations.

Multiblock high order Large Eddy Simulation of powered Fontan hemodynamics: Towards computational surgery

Children born with only one functional ventricle must typically undergo a series of three surgeries to obtain the so-called Fontan circulation in which the blood coming from the body passively flows from the Vena Cavae (VCs) to the Pulmonary Arteries (PAs) through the Total Cavopulmonary Connection (TCPC). The circulation is inherently inefficient due to the lack of a subpulmonary ventricle. Survivors face the risk of circulatory sequelae and eventual failure for the duration of their lives. Current efforts are focused on improving the outcomes of Fontan palliation, either passively by optimizing the TCPC, or actively by using mechanical support. We are working on a chronic implant that would be placed at the junction of the TCPC, and would provide the necessary pressure augmentation to re-establish a circulation that recapitulates a normal two-ventricle circulation. This implant is based on the Von Karman viscous pump and consists of a vaned impeller that rotates inside the TCPC. To evaluate the performance of such a device, and to study the flow features induced by the presence of the pump, Computational Fluid Dynamics (CFD) is used. CFD has become an important tool to understand hemodynamics owing to the possibility of simulating quickly a large number of designs and flow conditions without any harm for patients. The transitional and unsteady nature of the flow can make accurate simulations challenging. We developed and in-house high order Large Eddy Simulation (LES) solver coupled to a recent Immersed Boundary Method (IBM) to handle complex geometries. Multiblock capability is added to the solver to allow for efficient simulations of complex patient specific geometries. Blood simulations are performed in a complex patient specific TCPC geometry. In this study, simulations without mechanical assist are performed, as well as after virtual implantation of the temporary and chronic implants being developed. Instantaneous flow structures, hepatic factor distribution, and statistical data are presented for all three cases.