Biomedical Engineering Reference
In-Depth Information
or all of the following steps:
1.
Preprocess one of the motion-corrected raw functional images, cor-
recting for distortion (if not already done), removing bias field
and
or nonbrain structures
2.
Preprocess a structural image (of the same subject) to remove bias
field and
or nonbrain structures
3.
Register the raw functional image with the structural image
4.
Apply the transformation found in the previous step to the activa-
tion image
Note that the raw functional image is registered to the structural image
rather than the activation image itself. This is because the activation image
only contains color “blobs” of activation (see Color Figure 8.6) which do not
resemble a brain, so it is unsuitable for direct registration. However, since it
shares the same coordinate system as the raw functional image, it can be
transformed to the new coordinate system using the same transformation as
estimated for the raw functional image.
This registration is intrasubject but intermodality. Therefore, a low DOF
transformation (e.g., affine or affine plus restricted nonlinear registration,
such as a few low frequency spatial basis functions) can normally be used,
and is in fact desirable, since the resolution and contrast are poor in the EPI
images, making them unsuitable for less restrained high DOF nonlinear
registrations.
Figure 8.6 shows the results of an experiment in which both visual and
auditory stimulation were applied. Because the two stimulation types
were applied with different timings, statistical analysis can separate acti-
vation due to visual stimulation (red-yellow areas) from activation due to
auditory stimulation (blue). The top row shows several consecutive slices
from the original data with activation overlaid. This is the original resolu-
tion of the fMRI data. However, since the anatomical detail in the raw
functional images is poor, it is useful to render the activation (resampled
to high resolution after registration, as described above) onto the subject's
high resolution structural image. This greatly assists in locating the pre-
cise anatomical regions of the activations, as shown in two example slices
in the bottom row of the figure.
8.4.2
Functional to Standard Template
Another common application of registration in fMRI is the registering of
activation images with a template in order to assign standard coordinates.
This topic is treated in considerable detail in Chapter 14. These standard
coordinates are particularly useful, as they allow the spatial locations of
the activations to be reported and interpreted in a consistent, objective
way. (An example of a standard coordinate system is the one proposed by
