Biomedical Engineering Reference
In-Depth Information
Figure 12.
Illustration of the deformation maps involved in the proposed approach.
ϕ
f
is the map from the atlas to a subject's tumor-bearing image. Regions
T
SD
and
D
SD
denote the bulk tumor and edema regions in the subject's images, and
T
A
,D
A
are the
corresponding regions in the atlas.
ϕ
c
is the mapping from the atlas to the subject's image
before tumor mass-effect simulation (
B
S
is not known for non-simulated cases), and
ϕ
d
is that obtained through simulation of the tumor mass-effect. Simulating the tumor mass-
effect on the atlas results in
ϕ
a
and a deformed atlas image that can then be registered to
the deformed subject's image through
ϕ
b
. Copyright c
2006, Elsevier B. V. See attached
CD for color version.
approach is to find the transformation
ϕ
f
:
B
A
\
T
A
→
B
SD
\
T
SD
(1)
that maps points with coordinates
X
A
in the atlas image to points with coordinates
X
SD
in the subject image. Here, the
is used to denote the set difference operator.
Another goal of the deformable registration approach is to identify
T
A
, which
corresponds to brain tissue that is no longer present in the subject's image (died
or invaded by tumor).
As will be illustrated with an example later in Section 4.5, the direct use of
deformable image registration to obtain
ϕ
f
results in an inaccurate warping in and
around tumor due to a substantial dissimilarity between the normal atlas image
and the patient's tumor-bearing image. The approach proposed here precedes this
deformable registration step with a simulation of the tumor mass-effect on the atlas
image.
In Section 4.3 a biomechanical FEmodel for the deformation induced by brain
tumors is described and partially validated via real serial tumor scans. This model
\
