-- J.S. --
> I am trying to normalize a 14 slice functional volume, that is 70 mm,
> acquired in the axial plane. I have a 28 slice full brain anatomic volume
> which I have coregistered the mean of the functional volume to. This I
> used to determine the parameters for normalization. However the images
> that show up in the display after determining the parameters do not occupy
> the whole brain volume as the template does.
> With coregistration
> TARGET=mean functional image (14 slice)
> OBJECT=anatomical volume (28 slice)
>
> Is this correct?
The brain coverage in your image is not as big as the coverage of the
bounding box specified for writing the normalized data. If this data
is not present in the original image, then it can not be included in
the normalized image. This is pretty normal.
-- J.S. --
> If so are the normalized parameters enough to normalize the functional
> volume?
> Can these parameters be used to normalize the 11 functional runs from this
> study session?
> Or does each run have to be coregistered with the corresponding mean
> image?
The spatial normalization parameters can be applied to the functional images
providing they are all in register with the image that you estimated the
spatial normalization parameters from.
-- J.S. --
> Can the normalization parameters be determined from the 28 slice anatomic
> volume and applied to the 14 slice functional volume?
Yes, providing that the images have been co-registered. It is preferable to
estimate the spatial normalization parameters from the original structural
image after coregistration (rather than the resliced version) - since the
resolution is preserved and there is no missing data (as there often is at
the edges of resliced images). The information in the .mat file that is
created for the structural image during co-registration shows how the
orientation of the structural image relates to the orientation of the
functional image (at the time of co-registration).
Providing that both the coregistration of the structural image to the
functional mean, and the spatial normalization of the structural image
to the template both worked, then the normalization parameters should
work for your functional images.
One problem that the coregistration step can not currently overcome is the
distortion of the functional images. It assumes that the images can be
registered via a rigid body transformation (three rotations and three
translations) - which can not model non-linear distortions. If the distortions
can be corrected, then the coregistration will work better, and youwill get
better spatial normalizations of your functional images.
-- M.R. --
> in your case I would use the structural scan to derive the
> normalization parameters. The reason to do so, is that this scan coveres
> the whole brain and can be matched well to the template, while your
> functional scans cover only 70 mm. Otherwise you have to retrict the
> normalization parameters.
This is very true. However, retricting the normalization parameters should
be less important with SPM99.
-- M.R. --
> If the normalization of the structural scan failes, you should inspect the
> ORIGIN field in the header file. It should point to the location which
> corresponds to the ORIGIN of the template.
Exactly. However, another similar reason for a failure would be that the
".mat" file of the structural image is messed up. If an image has a ".mat"
file, then the ORIGIN information is extracted from it (rather than using
the ".hdr" file).
-- J.S. --
> If I choose to determine the parameters for normalization from the
> full brain anatomical T1 images, does any realignment need to be done to
> this T1 volume before parameters are determined?
Just the coregistration of the structural to the functionals mentioned
earlier.
Good luck,
-John
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|