Image Processing Reference
In-Depth Information
the data clinically available. Therefore, clinical pilot studies have been possible and
have shown the relevance and potential of this data. Furthermore, theMRI acquisition
development towards 7 and 9T machines have the potential to provide the required
resolution and signal-to-noise ratios (SNR) to analyze blood flow in smaller vessels
compared to the main arteries around the heart.
In this chapter, we will consider the visualization challenges for both simula-
tions and in-vivo measurements of unsteady flow fields. We will present a review of
the existing literature, the main challenges related to blood flow visualization and
analysis, as well as the open issues.
25.2 Blood Flow Simulation
One way of determining blood flow within a vascular system is through simulation.
This approach involves two major steps. First, the vascular structure needs to be
segmented, and the geometry of the vessel boundary determined as accurately as
possible. Next, a Computational Fluid Dynamics (CFD) model simulates the blood
flow within the reconstructed geometry. The next sections will discuss these steps in
more detail.
25.2.1 Grid Generation
In order to simulate blood flow with CFD, the boundary conditions of the underly-
ing mathematical model have to be defined properly. In case of vascular flow, the
boundary conditions are defined by two different components: the first one is the the
geometric boundary of the vessels; the second one consists of the inflow and outflow
characteristics as defined by the circulatory system.
There are different ways of identifying the vessel boundary. Typically, some imag-
ing technique is used to generate a scan of the vascular structure for which the flow is
supposed to be simulated, for example a Computed Tomography (CT) scan. In order
to extract the vascular structure from such a volumetric image, the data needs to be
segmented. Simple thresholding based on the intensity value can be used. However,
this may not be sufficient for anatomical structures where significant perfusion and
noise occurs, such as the heart. More sophisticated segmentation techniques are nec-
essary, for example, gradient-based thresholding techniques tend to produce better
results in those cases.
The segmentation process also has great influence on the overall accuracy of the
simulation. Basic intensity thresholding techniques, for example, determine individ-
ual voxel locations as being part of the vessel boundary. However, it is unlikely that
the vessel boundary is located precisely at such a voxel location, especially given that
the volumetric data set only imposes an artificial grid on the organ at hand. There-
Search WWH ::
Custom Search