Biomedical Engineering Reference
In-Depth Information
pressure. In the current formulation, we can recover the inlet/outlet flow delay by
setting the outlet pressure boundary conditions at the descending aorta proportional
to the flow, as measured by PC-MRI on the descending aortic cross-section.
A possible extension of our current setup is the inclusion of turbulence
modeling. While many previous works have not included turbulence shear terms
in their models, such terms may be needed due to the high Reynolds numbers
(sometimes greater than 8,000) achieved during systole in the AAo, especially
considering that we are using pathological cases which are often characterized
by increased Reynolds numbers. In our framework, a first goal is to validate
qualitatively the patterns obtained using CFD and the ones measured by PC-MRI.
Due to the measurement location in the descending aorta being far from the
regions in the AAo where turbulence is generated, we chose to start with the
nonturbulent model. This allows us to better quantify the influence of the aortic
geometry on the flow.
3 Experiments and Validation
This section presents results of hemodynamics simulations performed with the
proposed method on randomly selected patients from the COAST [
20
]. COAST
is an FDA-sponsored multicenter clinical trial that investigates the safety and
effectiveness of Cheatham Platinum (CP) bare metal stents in the treatment of
native and recurrent aortic coarctation. COAST is a multicenter trial, involving
patients enrolled from multiple pediatric institutions across the United State, with
data acquired on MR scanners from multiple vendors by various radiologists,
employing varying imaging protocols and different parameter encoding. All images
used within our evaluations are acquired during follow-up studies at 12-month after
transcatheter implantation of metallic stents.
Using the CFD simulation framework described above, we have performed a
series of simulations using the geometric constraints of the aortic meshes as static
boundary conditions, and the sampled MRI-derived velocity as the inflow profiles
(see Fig.
2
). The aortic data was randomly sampled from patients with various
pathologies, including bicuspid valve, coarctations, artificial valves, and stents. The
full comparison of the simulations is beyond the scope of this (system) paper, and
we give here an outline of several observed patterns that correlate with the various
pathologies.
The blood density and dynamic viscosity used in the simulations were set
to generic mean values across healthy individuals, namely
cm
3
r
¼
1
:
05
g
=
and
m
¼
s
. The simulations featured grids with physical resolutions between
0.5-1.5 mm, and were run on Windows PCs with Intel Xeon CPUs of 2.53 GHz.
Two cycles were computed in each experiment, with total computation times
varying between 12-18 hours. The multigrid PCG solver used to iteratively
solve the pressure Poisson equation and the Poisson equation arising from the
4
mPa
