Information Technology Reference
In-Depth Information
been proposed. By generating oblique sagittal images, Rabassa et al. [
62
] showed
that visualization of vertebral facet joints improved, while oblique axial images
allowed views that were parallel to intervertebral discs. Although the visualization
was limited to oblique cross-sections, the authors concluded that in certain clinical
situations, such as for the evaluation of neural foraminal stenosis or localization of
spinal lesions, reformatted images could supplement the original 3D images.
Oblique cross-sections that were orthogonal to the long axis of both left and right
neural foraminae of the cervical spine region were also generated by Roberts et al.
[
65
], who showed that by oblique MPR, consistency in the interpretation of neural
foraminal stenosis between observers was improved, and suggested that such an
approach should be considered in routine evaluation. Rothman et al. [
68
] demon-
strated that curved cross-sections, obtained by connecting manually selected points
into a continuous curve, were useful for the evaluation of anatomical relationships
in the coronal spine region. After reformation, structures such as nerve roots,
vertebral facet joints and spinal cord could be observed in a single 2D cross-section.
Congenital spinal abnormalities were examined by Newton et al. [
50
], who man-
ually outlined the boundaries of the spine in multi-planar cross-sections and created
curved cross-sections that improved the identi
cation and interpretation of abnor-
malities. The bene
t of curved cross-sections was, in comparison with multi-planar
or oblique cross-sections, most valuable in the case of signi
cant sagittal or coronal
spinal curvature, as they may help spine surgeons to achieve a more complete
understanding and evaluation of spinal deformities. Menten et al. [
48
] presented a
curved planospheric reformation method that was based on the reconstruction from
a cylindrical plane, de
ned around the approximate boundary of the spinal canal
within an axial cross-section. As a result, the anterior and posterior anatomical
structures of the spine were displayed simultaneously in the same plane, which
improved the evaluation of congenital spinal deformities. Manual determination of
points or curves that determined the curved cross-sections was required in all of the
above mentioned studies. A semi-automated method was presented by Kaminsky
et al. [
33
], who segmented the spine on reformatted 3D images in order to over-
come the problems of orientation in the standard multi-planar con
guration. The
transformation axis was determined by a 3D spline, obtained either manually by
delineating centerlines in sagittal and coronal cross-sections, or automatically by
dropping spheres of maximum possible radius through vertebral bodies or the spinal
canal. Vrtovec et al. [
83
,
85
] extracted curved cross-section from 3D spine images
by representing the spine curve and the rotation of vertebrae as polynomial func-
tions in 3D that formed the transformation axes for the reformation procedure,
while Klinder et al. [
36
] reformatted 3D images by stacking curved cross-sections
in order to reduce the region of interest and make the subsequent detection of
vertebrae independent of the spinal curvature. Hanaoka et al. [
22
] extracted curved
cross-sections by simultaneously aligning one elliptical column to the vertebral
bodies and intervertebral discs, and a second elliptical column to the spinal canal,
which allowed virtual straightening of the 3D image.
Image reformation was also identi
ed as a valuable visualization technique in
MR imaging of the spine. Apicella and Mirowitz [
4
] reported that multi-planar
