Objective: This study aimed to develop an accessible patient-specific computational flow modelling pipeline for evaluating the hemodynamic performance of fenestrated endovascular aneurysm repair (fEVAR), with the hypothesis that computational flow modelling can detect aortic branch hemodynamic changes associated with fEVAR graft implantation.
Methods: Patients who underwent fEVAR for juxtarenal aortic aneurysms with the Cook ZFEN were retrospectively selected. Using open-source SimVascular software, preoperative and postoperative visceral aortic anatomy was manually segmented from computed tomography angiograms. Three-dimensional geometric models were then discretized into tetrahedral finite element meshes. Patient-specific pulsatile in-flow conditions were derived from known supraceliac aortic flow waveforms and adjusted for patient body surface area, average resting heart rate, and blood pressure. Outlet boundary conditions consisted of three-element Windkessel models approximated from physiologic flow splits. Rigid wall flow simulations were then performed on preoperative and postoperative models with the same inflow and outflow conditions. We used SimVascular's incompressible Navier-Stokes solver to perform blood flow simulations on a cluster using 72 cores.
Results: Preoperative and postoperative flow simulations were performed for 10 patients undergoing fEVAR with a total of 30 target vessels (20 renal stents, 10 mesenteric scallops). Postoperative models required a higher mean number of mesh elements to reach mesh convergence (3.2 ± 1.8 × 106 vs 2.6 ± 1.1 × 106; P = .005) with a longer mean computational time (10.3 ± 6.3 hours vs 7.8 ± 3.5 hours; P = .04) compared with preoperative models. fEVAR was associated with small but statistically significant increases in mean peak proximal aortic arterial pressure (140.3 ± 11.0 mm Hg vs 136.9 ± 8.7 mm Hg; P = .02) and peak renal artery pressure (131.6 ± 14.8 mm Hg vs 128.9 ± 11.8 mm Hg; P = .04) compared with preoperative simulations. No differences were observed in peak pressure in the celiac, superior mesenteric, or distal aortic arteries (P = .17-.96). When measuring blood flow, the only observed difference was an increase in peak renal flow rate after fEVAR (17.5 ± 3.8 mL/s vs 16.9 ± 3.5 mL/s;P =.04). fEVAR was not associated with changes in the mean pressure or the mean flow rate in the celiac, superior mesenteric, or renal arteries (P = .06-.98). Stenting of the renal arteries did not induce significant changes time-averaged wall shear stress in the proximal renal artery (23.4 ± 8.1 dynes/cm2 vs 23.2 ± 8.4 dynes/cm2; P = .98) or distal renal artery (32.7 ± 13.9 dynes/cm2 vs 29.6 ± 11.8 dynes/cm2; P = .23). In addition, computational visualization of crosssectional velocity profiles revealed low flow disturbances associated with protrusion of renal graft fabric into the aortic lumen.
Conclusions: In a pilot study involving a selective cohort of patients who underwent uncomplicated fEVAR, patient-specific flow modelling was a feasible method for assessing the hemodynamic performance of various two-vessel fenestrated device configurations and revealed subtle differences in computationally derived peak branch pressure and blood flow rates. Structural changes in aortic flow geometry after fEVAR do not seem to affect computationally estimated renovisceral branch perfusion or wall shear stress adversely. Additional studies with invasive angiography or phase contrast magnetic resonance imaging are required to clinically validate these findings.
Keywords: CFD; Computational fluid dynamics; Fenestrated endovascular aneurysm repair; Patient-specific; fEVAR; thrombosis.