Visual Exploration of Simulated and Measured Blood Flow - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 290

Fig. 25.1 An example of a vessel boundary based on a CT scan segmented using gradient-based
thresholding with sub-voxel precision ( a ) and resulting simulation of blood flow visualized by
color-coding the wall-shear stress ( b )
fore, the accuracy can be improved by using segmentation techniques that operate at
a sub-voxel level, e.g., Marching Cubes algorithm [ 25 ].
The downside of the Marching Cubes algorithm, however, is that it only operates
with a fixed intensity threshold across the entire data set. Since typically a point
spread function with a radius greater than one has to be assumed for most imaging
techniques, this may cause errors in the boundary geometry, overestimating larger
vessels and underestimating smaller vessels. Gradient-based approaches can achieve
better results in these cases identifying the location where the maximal gradient
value is assumed to find a more accurate estimate for the exact location of the vessel
boundary [ 52 ]. The geometric model resulting from the segmentation step can then
be further refined, for example by using smoothing or rounding off the transitions at
vessel bifurcations [ 33 ], resulting in a vessel boundary that can be used for a CFD
simulation. Figure 25.1 a shows an example of such a vessel boundary generated
based on a CT scan using gradient-based thresholding with sub-voxel precision.
25.2.2 Computational Fluid Dynamics Model
In addition to the geometric boundary of the vascular structure inflow, outflow, and
wall boundary conditions have to be defined properly for the CFD simulation [ 5 , 49 ].
Inflow and outflow conditions arise from the fact that the current vascular structure
has to be isolated from the rest of the arterial system. In practice, the flow rate or
the speed profile at the inlets and the pressure at the outlets are utilized for a car-
diac cycle. These quantities are obtained based on experimental measurements or by
time-resolved Phase-contrast MRI flow measuring from the patient (see subsequent
section). boundary condition arises from the fact that the vessel wall is distensible,
which may influence the local hemodynamic and vice versa [ 44 ]. However, typi-
cally no proper characterization of the arterial wall, such as modulus of elasticity,
wall thickness, or pressure wave that form at the wall, is available or is difficult to
measure noninvasively [ 5 ]. Thus, a rigid wall is assumed in most cases, which also