Biomedical Engineering Reference
In-Depth Information
FIGURE 9.6
Result from MRI-PET registration employing automated and interactive techniques. Top
row: PET-FDG study with overlaid contours derived from MRI (bottom) after interactive
realignment. Middle row: result from a O-15-water PET activation study shown within the
contours of the PET-FDG study after automated registration via similarity measures. Bottom
row: MRI study with overlaid contours derived from activation study after interactive
registration via mean image from activation study.
5
of both modalities independently.
Typical examples in which different levels
of technical effort have been employed for comparing images are listed below,
classified according to the definitions outlined in Section 9.5.1.
Basic combination or correlation has been performed by Braams et al.,
3
15
16
Laubenbacher et al.
for extracranial head and neck carcino-
mas. The authors pointed out the definite advantage in staging the tumors
although no full spatial registration was performed. In spite of the unavailabil-
ity of full integration and fusion, simple visual comparison was found to be of
benefit for patient treatment. Ratib et al.
and Jabour et al.
17
focused on cardiac images. They did
not perform exact registration, but argued that the generation of standard
views based on planes oriented according to the cardiac geometry can provide
a practical means to compare the images with reasonable accuracy.
Only a few groups have dealt with the problem of exact registration, (i.e., pre-
cise computerized positional matching) of PET and MRI extracranial images.
Compared to neurological imaging of the brain, the exact registration of cardiac
images is hampered because the heart is a nonrigid moving organ. Possible pro-
cedures for multimodal cardiac imaging have been reviewed by Gilardi et al.,
18
19
among them a protocol suggested by Sinha et al.,
who performed PET and
