This subproject is one of many research subprojects utilizing theresources provided by a Center grant funded by NIH/NCRR. The subproject andinvestigator (PI) may have received primary funding from another NIH source,and thus could be represented in other CRISP entries. The institution listed isfor the Center, which is not necessarily the institution for the investigator.Abdominal aortic aneurysms (AAAs) are an increasingly prevalent cardiovascular disease characterized by a dilation of the infrarenal aortic segment. Currently AAAs are estimated to affect about 2 million Americans; within men aged 65- to 75-years old the incidence rate is about 3-5% and likelihood 10 times greater than similarly aged women. While their exact mechanism for development is unknown, they are associated with a cumulative effect initiated by the altered flow patterns within the aorta near the bifurcation of the iliac arteries. As an aneurysm develops, there exists a changing dynamics and mechanics that create localized stress gradients; these gradients are predicted to further contribute to the weakening of the arterial wall until the point where mechanical failure occurs. Failure of the aneurysm wall is characterized as aneurysm rupture, an event associated with a 90% overall mortality rate, representing the 13th leading cause of death in the U.S. While AAAs remain largely asymptomatic, they are treatable with surgical intervention once properly diagnosed. One technique used to restore normal blood flow through the aneurysm and reduce the pressure exerted along the wall is referred to as endovascular aneurysm repair (EVAR). It is a minimally invasive procedure in which an endovascular graft (EVG) is placed within the aneurysm through a small incision in the femoral artery. These EVGs are then expanded and attached either by an active or passive mechanism to the arterial wall. An active mechanism requires the use of stents, barbs, or hooks at the proximal and distal ends, which are fixed into the wall of the aortic neck and iliac arteries. Alternatively, a passive system can be used in which an extension cuff acts to oversize the graft and hold it in place through a balance of frictional forces. While EVAR is a clinically advantageous technique, it demonstrates an increased rate of secondary intervention due to the development of endoleaks and graft migration; two events which are characteristic of EVG failure. Incidence rates for endoleaks vary between 5-36% while graft migration occurs in approximately 15% of the cases. These failure mechanisms repressurize the aneurysm sac, increasing its susceptibility to rupture. Displacement forces and stresses induced by blood flow through a repaired aneurysm are believed to contribute to EVG failure. The objective of this project is to quantify the hemodynamics under pre- and post-operative conditions and their effect on the luminal surface, either the diseased arterial wall or the endovascular graft. Specifically this project will examine the contribution of the interaction between the blood flow, arterial wall, and intraluminal thrombus (ILT) within patient-specific aneurysms prior to repair, with the inclusion of the EVG after EVAR. With the support of a PSC Production Grant under the Biomedical Supercomputing Initiative, fully-coupled fluid-structure interaction (FSI) simulations of reconstructed AAAs and EVGs can be used to perform a complete investigation on the factors and forces which contribute to graft failure. Our hypothesis is that an improved graft design can be developed from the results of large-scale computational fluid dynamics and solid mechanics analyses with experimental validation. From these patient-specific AAAs reconstructed from CT Scans obtained through the Vascular Biomechanics Research Lab of the University of Pittsburgh Medical Center (UPMC) the influence of (i) pulsatile blood flow, (ii) aneurismal wall properties, (iii) intraluminal thrombus, and (iv) endovascular graft attachment can be assessed under physiologically realistic conditions. Computed tomography (CT) scans performed at least 3 months prior to and 30 days after EVAR have been obtained from the database of the University of Pittsburgh Medical Center for three patients. These images were reconstructed using an edge-detection protocol to identify the specific regions of interest, namely the boundary of the arterial wall and luminal surface area. The difference between these areas represents the presence of thrombus. The smoothed spline contours generated from the reconstruction algorithm are imported into a NURBS software where cross-section contours are blended and modified to create the appropriate surfaces. These files are exported into IGES format which represent the input for our research. Various commercial software packages are used for mesh generation including ADINA (v.8.2, Cambridge, MA) and Gambit (v 2.2.30, Fluent, Inc., Lebanon, NH) for linear tetrahedral and hexahedral elements respectively. This is completed at the computational and experimental fluid mechanics lab on XP or Linux workstations. The fully-coupled fluid structure interaction simulations need to be completed on a computational platform with a significant memory allocation. The commercial software ADINA will be used, which uses the finite element method to solve the discretized fluid and solid domains. A progression of simulations is completed to study the independent effects of each domain before coupling the fluid-structure interaction. Initially rigid-wall fluid simulations are completed under steady and subsequently transient conditions using an applied inlet velocity and stress-free outflow conditions. Static and dynamics stress analyses are completed with a uniform applied pressure load generated from a physiological waveform. To gage the influence of fluid flow on the solid domain, a fully-coupled fluid-structure interaction is completed first under initial steady-state conditions of an initial velocity and outlet pressure. The results from this analysis are then used as a restart condition for the dynamic FSI cases. ADINA solves the fluid-structure interaction by incorporating the solutions of the Navier-Stokes equations for the fluid domain and dynamic equilibrium for the solid domain. The establishment of the fluid-structure boundary allows the coupling of these solutions so that the resulting velocity, pressure, and displacement of the fluid model are imposed on the solid model using an arbitrary-Lagrangian-Eulerian formulation (ALE) for the moving mesh. The resulting fluid stress and displacements determined from the solid domain are in-turn applied to the fluid boundaries once force equilibrium has been attained. Initial mesh comparison analysis have been completed which indicate a mesh density of over 165,000 nodes may be needed to attain convergence of secondary criteria important to this project for fluid analysis, including wall shear stress values. Within the ADINA code two types of solvers are suggested for fluid flow simulation: sparse and biconjugate gradient (BCG). A linear system of algebraic equations is formed during the discretization of the Navier-Stokes equations over each element. In order to decompose and solve this matrix either a direct (sparse) or iterative (BCG) solution technique is used, depending on the computational domain, predicted fluid flow behavior, and system requirements. Each solver technique initially stores the matrix before re-ordering or pre-conditioning the equations within the matrix for processing. The iterative method typically enjoys the advantage of reduced computational memory and computing time. However, there has been some discussion regarding its validity within fluid flow solutions. Running identical simulations under both solver techniques show the sparse solver requires nearly twice as much memory for matrix solution, nearly 3GB of memory for solution processing, and approximately 2.2GB for solution storage. The optimum HP workstation used at our facilities includes a dual-processor, 64-bit, 1.56GHz station with 8GB RAM. A transient dynamic analysis completed on this processor with tetrahedral elements using the BCG solver takes about 5,000 time steps to converge to a steady-state solution for a total run time of about 215 wall-clock hours. A sparse solver solution for the same problem may take over 30 times that long to process. Solid mechanics analyses are run with the sparse solver and may be processed at our facilities for static and dynamics simulation. However when a fully-coupled simulation is completed, the memory required takes into account both fluid and solid domains as well as the formulation for moving the mesh. Therefore the memory required to complete the simulation, coupled with the wall-clock hours, will be considerable and needs to be completed on a platform such as the one PSC offers. A hypothetical FSI case with approximately 100,000 nodes and 80,000 linear hexahedral elements has run on Rachel under steady state conditions. Using the sparse solver to solve an initial case approximately 10GB of total memory was required, along with 25 wall-clock hours to run 20 time steps and about 2500 SUs. A typical transient analysis would require a similar memory with approximately 5,000 time steps.
Showing the most recent 10 out of 292 publications
