Information Technology Reference
In-Depth Information
often unsatisfying, which may seriously affect the quality of the diagnostic infor-
mation of the observed curved structures. When visualizing 3D spine images with
multi-planar cross-sections, the spine may intersect with sagittal and coronal planes,
while the axial plane may not always be located at the same level of vertebral
bodies or intervertebral discs. The important structural parts of the spine may
therefore not be displayed simultaneously in any single multi-planar cross-section,
which may therefore not provide suf
cient or qualitative enough diagnostic infor-
mation, because they cannot follow the curvature of the spine and the rotation of
vertebrae. This is already the case when visualizing a normal spine due to its natural
“
-shaped curvature, and is even more emphasized in pathological spinal curva-
tures, for example in the case of scoliosis or increased kyphosis/lordosis. Curved
planar reformation (CPR) is a 2D image visualization technique that displays the
originally reconstructed pixels along any user-de
S
”
at-
tened in order to appear as a plane. The use of the CPR visualization technique,
which generates cross-sections that are orthogonal or tangent to the curve along the
structure, represents a solution to the above mentioned problem. The standard
coordinate system, which is determined by the 3D image, is transformed into a
coordinate system that is determined by the observed 3D anatomical structure, for
example, into the spine-based coordinate system in the case of 3D spine images
(Sect.
2.1.2
).
As a visualization technique, CPR is used in the field of angiography to display
and evaluate blood vessels [
25
,
34
,
35
,
46
,
51
,
61
,
63
,
64
,
72
], in the
ned curved surface that is
fl
eld of
pancreatography to display and evaluate pancreatic diseases [
21
,
60
], for brain
visualization [
43
], in the
field of colon-
oscopy [
19
,
71
,
92
]. In all of the CPR visualization approaches, the determination
of the curve that represents the central course of the visualized tubular structure is of
utmost importance [
3
,
5
,
9
,
41
,
97
]. Dedicated commercial software or software
provided by CT and MR scanner manufacturers already allows generation of curved
cross-sections, however, this requires manual determination of the curve that fol-
lows the anatomical structure. Although MR scanners allow arbitrary orientation of
the imaging plane and can therefore simulate the generation of oblique cross-
sections, such visualization is greatly in
field of bronchoscopy [
42
,
57
] and in the
uenced by the scanner operator that
determines the orientation of the imaging plane and by the position of the subject in
the scanner. Curved cross-sections can be also acquired directly from the MR
scanner [
10
,
11
,
28
], however, the quality of the obtained images in not adequate
due to low spatial resolution of images, presence of intensity modulation artefacts
and the fact that images can be curved only in one dimension. New concepts in
curved-slice imaging allow to maintain a close-to-rectangular voxel size [
93
] and
constant cross-sectional thickness [
94
]. On the other hand, by applying a combi-
nation of linear and/or non-linear spatial encoding magnetic
fl
fields for excitation and
geometrically matched local encoding of curved-slice imaging, it is possible to
achieve an almost rectangular voxel size [
93
] and constant cross-sectional thickness
[
94
] of curved cross-sections.
Many approaches that aim to improve quantitative and qualitative evaluation of
spinal deformities by an effective visualization of CT spine images have already
