Biomedical Engineering Reference
In-Depth Information
FIGURE 15.4
Examples of an experiment in progress with a fluoroscopic film illustrating a complete implant.
have been used, including a temporally mounted piston
32
and temporally and
intraparenchymally located balloon catheters.
33
These deformation sources
were moved incrementally in a controlled fashion with full CT image acquisi-
tions occurring between each intervention event. Rigid-body image registra-
tion was maintained from scan to scan by the stereotactic frame which was in
place and the fact that the animal's position was not changed once inside the
CT scanner.
Interestingly, this group chose to use MR images to define their computa-
tional model, but CT scans to track deformation events. MR scanning for
model construction is a logical choice because of its superior soft tissue contrast,
allowing computational models which differentiate gray and white matter to
be employed. In terms of deploying the CT for deformation tracking, they have
clearly traded off the error induced by having to coregister the model (con-
structed from MR) with the CT images obtained during an experiment with
the fast image acquisition and artifact reduction (around the small metallic
tissue markers) afforded by the CT. The investigators have quantified the
impact of registration error on model validation by randomly perturbing the
starting location of each bead in the model (i.e., the uncertainty of bead
location resulting from potential registration error) up to 5 mm, and con-
cluded that model comparisons vary no more than 10%. Figure 15.5 shows a
