Biomedical Engineering Reference
In-Depth Information
As discussed in Chapter 13, the deformation field produced by nonrigid intra-
subject registration algorithms can quantify normal development and contribute
to an understanding of disease processes and aging. The nonrigid registration
algorithms used in these applications will soon be reliable enough to enter clinical
use, providing valuable tools for diagnosis and monitoring disease progression.
16.2.4
Group Differences and Postgenomic Registration
When nonrigid intersubject registration algorithms are applied to cohorts of
patients and matched normals, the calculated deformation field has the
potential to provide a very sensitive measure of structural differences
between the groups. Much work has already been done in the study of
schizophrenia, but deformation fields might reveal effects in other disorders
that interfere with structure. This should contribute to a better understanding
of disease and perhaps even provide a link between genes, structure, and
function. Integrating imaging information obtained
may provide
direct and powerful insights into gene function and genetic control in whole
functioning organisms as opposed to model systems or
in vivo
specimens.
Postgenomic registration studies are beginning to be applied to the study of
gene expression in mice, and intersubject registration of human subjects cor-
related with sequenced genomes could be even more revealing. This is likely
to be a key tool in studying gene-environment interactions.
ex vivo
16.2.5
Registration for and Combined with Image Segmentation
The difficult problem of image segmentation has traditionally been thought of
as quite a different research topic to image registration, but several groups have
now demonstrated that good segmentation can be achieved by lining up a sub-
ject's images to an atlas using a nonrigid registration algorithm (e.g., Dawant
et al.
1
). Labeled structures in the atlas can then be used to split up the images
into anatomical and pathological components for visualization or quantifica-
tion. A related use of image registration is to fuse functional and metabolic
information obtained by methods that reveal little of the structural information
onto segmented anatomical images, or vice versa. For example, high resolution
data segmented by tissue classification may allow much more subtle changes
in metabolism or perfusion to be detected. Several applications of this general
type have been discussed in this topic, but further developments and wide-
spread use are likely to allow much more effective use of the data.
16.2.6
Registration to Improve Image Acquisition
Registration is beginning to be used to improve image acquisition. For exam-
ple, on-the-fly registration can be used to dynamically adjust slice position in
MR scans to compensate prospectively for subject motion (e.g., Thesen et al.
2
)
Registration of previously acquired images could also be used to reduce the
