Biomedical Engineering Reference
In-Depth Information
with
R
=
R
X
R
Y
R
Z
is the combination of rotations around the three orthogonal
axis (
X
,
Y
,
Z
) defining the 3D domain, and
T
=
[
T
X
,
T
Y
,
T
Z
] is the translation
vector in each axis direction.
The function
f
(
x
) is defined in a homogeneity-based framework [40] as:
f
(
x
)
=
f
in
(
x
)
−
f
out
(
x
)
.
(2.39)
The functions
f
in
and
f
out
are defined as:
f
in
=
(
I
−
u
)
2
f
out
=
(
I
−
v
)
2
(2.40)
where
u
and
v
denote the mean values of the image
I
inside and outside the
surface
S
. Analogous definitions of
f
with statistics on
I
are also derived.
The authors reported three experiments on simultaneous segmentation and
registration of MRI/CT images of the head and the spine both in 2D and 3D.
Validation via visual inspection showed accurate contour extraction for these
limited experiments.
2.4
Review of Clinical Validations
In this section we review in detail several recent papers that apply level set
segmentation and registration methods to medical images and provide a detailed
validation of their method through a clinical study for qualitative and quantitative
assessment of the accuracy of the method in assisting or performing a particular
clinical diagnosis task.
2.4.1
Important Clinical Segmentation Problems
We introduce in some details the two major applications in the domain of seg-
mentation of clinical images: Segmentation of the brain and segmentation of the
left ventricular cardiac cavity.
2.4.1.1
Segmentation of Brain Images
The two major modalities used for brain screening are MRI and SPECT/PET.
