Biomedical Engineering Reference
In-Depth Information
entered clinical protocols to support radiation therapy or special surgical
interventions with complementary PET/MRI information. With interactive
registration techniques and automated methods based on appropriate algo-
rithms, even PET images obtained with special tracers like [
11
C]-methionine,
which contain little anatomical detail, can be combined with MRI for clinical
diagnosis or treatment planning. There are, however, many potential difficul-
ties in the registration of MR and PET brain images, and this chapter has
emphasized the importance of careful image acquisition as well as the use of
a suitable algorithm for aligning the images.
For extracranial studies, exact registration is frequently hampered by the fact
that many current algorithms employ rigid-body transformations, which can
fail when organs move between measurements or because of cardiac or respira-
tory motion. However, registration of clinically acceptable accuracy seems fea-
sible when restricted to a relatively small coverage of the body, like thorax
studies. For head and neck studies, with careful positioning in both modalities,
registration looks promising. PET transmission imaging has an interesting role
as a link between PET emission and MR images. Little effort is involved in
obtaining transmission images, since their information is needed for attenua-
tion correction of the emission images. Recent progress in nonrigid registration
algorithms, as discussed in Chapters 13 through 15 of this topic, may lead to the
availability of better algorithms for extracranial MR-PET registration. Such
progress in combination with increasing availability of both PET and MRI in
clinical settings is likely to lead to widespread application of medical image reg-
istration as a vital link in maximizing the diagnostic value of these complemen-
tary modalities in the future. In the scientific realm, the improvements in
sensitivity and resolution that have been achieved for PET imaging have led to
widespread use of image registration as a core component in fusing functional
and anatomical data in single subjects. This increases study power by explicitly
accommodating individual anatomical variability directly in functional studies.
Increasing availability and reductions in costs of the computing facilities
required for image registration are also helping to stimulate more extensive
use of the technique. A critical issue in all clinical applications is the valida-
tion of proposed techniques and a clear demonstration of their utility, effi-
ciency, and cost effectiveness. The past two decades have seen extensive
development of and experimentation with candidate methods and applica-
tions. The next phase of development is likely to include commercialization
of standard packages and larger scale studies to provide the evidence neces-
sary for their adoption in clinical practice.
Acknowledgments
The author is indebted for many helpful discussions with A. Thiel, M.D., Pro-
fessor K. Herholz, M.D., W. Eschner, Ph.D., K. Scheidhauer, M.D., P. Theissen,
M.D., F. Maes, Ph.D., Professor R. Chisin, M.D., and to H. Herzog, Ph.D., for
carefully reading of this manuscript.
