Biomedical Engineering Reference
In-Depth Information
to the imaging device* include physical or technical limitations of the imag-
ing process and suboptimal or variable performance levels. Interactions
between imaged object and imaging device, on the other hand, constitute an
inherent part of the imaging process and also often lead to artifacts. For
example, the imaging process may be optimized to capture accurately one
specific component or aspect of the imaged object to the detriment of another,
therefore leading to artifacts (e.g., chemical shift artifact in magnetic reso-
nance imaging). Furthermore, artifacts resulting from the interaction
between object and imaging device may also be unavoidable in certain cir-
cumstances (e.g., the streak artifact in CT; see Joseph
1
).
This chapter is concerned primarily with the issue of correcting image arti-
facts due to scanner errors, i.e., mechanisms intrinsic to the scanner or the
imaging process itself and, to a lesser extent, consideration of artifacts intrin-
sic to the imaged object.
5.2
Geometric Distortion in Computed Tomography
CT is the oldest 3D medical imaging modality and has probably been the sub-
ject of the largest amount of technical development and refinements among all
the modalities considered here. However, there is relatively little published
work on the issue of CT geometric distortions, particularly in the context of
image registration. This may be because CT is based on the absorption of radi-
ation that traverses the body in a straight line, resulting in a very low degree of
spatial distortion. Most artifacts in CT concern intensity, in particular the streak
artifacts which have a variety of sources (see Joseph
1
). However, we will not
discuss these in detail since most of these are either unavoidable, particularly
and most importantly streaking due to the presence of metalic objects (e.g., ste-
reotaxic frame) in the imaged volume, or do not constitute a significant prob-
lem for registration of images from modern scanners.
CT is closely linked to the development of frame-based stereotaxy for local-
ization and co-registration; see Kelly and Kall.
2
Frame-based stereotaxy, by
providing an external coordinate system in each slice, insures a high degree of
robustness to images relative to scanner errors.
3
However, the advent of frame-
less stereotaxy, discussed in detail in Chapter 12, requires a reassessment of the
sources of distortion and possible new methods to minimize these will need to
be devised. Nonetheless, CT is generally assumed not to necessitate distortion
correction in contrast to MRI.
4,5
Studies on the accuracy of localization in CT
based on stereotactic frames have consistently reflected the low degree of geo-
metric distortion in CT, with in-plane errors less than 1 mm and effectively lim-
ited by voxel dimensions, in particular slice thickness and separation.
6-9
* We can also put into this category the factors related to the external environment (e.g., ambient
conditions that fall outside the specified operational range of the device).
