Abstract
Hemodynamics of endovascularly treated intracranial aneurysms has been proven to be one of the essential mechanisms driving treatment success due to its intimate relationship with processes involved in thrombosis and leadingto a stable clot inside the aneurysm. Giving access to hemodynamics, patient-specific computational fluid dynamics(CFD) simulations have been actively performed during the last decade in an attempt to both enhance existing devicesand predict the chances of success beforehand to the surgical act. Nevertheless, “classical” CFD simulations, referredto as conformal, yield high computational and meshing costs due to the heterogeneity of length scales between thedense weave of the fine struts of the device and the arterial volume. Homogeneous strategies recently developed tocircumvent this issue substitute local dissipations due to the wires with a global effect in the form of a pressure-dropacross the device surface. However, these methods cannot accurately reproduce the flow-patterns encountered near thestruts, despite the fact that the latter strongly dictates the downstream intra-saccular flow environment.This thesis aims at developing a computational model correctly reproducing local wires-induced flow heterogeneitieswhile keeping memory consumption, meshing and computational times as low as possible. A framework based on theImmersed Boundary Method (IBM) is introduced and validated on both idealized and patient-specific geometriestreated with endovascular devices. It is shown that the present model compares qualitatively and quantitatively wellwith conformal results for flow-diverters (FDs) but more importantly, it yields results that are either comparable orbetter than homogeneous methods with gains of one and three orders of magnitude for memory and computationaltime compared to conformal, respectively. Moreover, the proposed approach has proven its versatility to correctlyaccount for other braided endovascular devices such as intra-saccular WEBs.Finally, a database of numerical computations using the present model has been built using 27 patient-specificgeometries treated with WEBs and for which the treatment outcome is known. Semi-automated numerical tools usedto build this database and intended to non-CFD specialists are presented. Preliminary results from a study dedicatedto treatment outcome prediction with both geometrical and hemodynamics indices derived from this database aregiven and discussed. Several important limitations are drawn and should be considered in future works.