Frederick,
I was running into similar problems for a couple of my brains. The
problem was that the specified origin of the volume that I was trying to
reconstruct was too far from the midline. I'm attaching some notes that I
wrote while solving the problem:
Notes on getting Freesurfer to work and aligning data with SPM.
To get the origin of the hires set properly, bypass the csurf
interface and use mri_convert directly. Example:
1. Use GE2SPM to create analyze format file.
2. Determine voxel offset of AC from origin of the analyze format file
which should be 62 128 128.
3. mri_convert -ic 62 128 128 -oc 61 167 144 -iid -1 0 0 -ijd 0 1 0
-ikd 0 0 1 -ot cor /data2/modfmri/28jan02MC/hires/28jan02MC_002.img
/data2/modfmri/28jan02MC/mri/orig/
where -ic are the voxel coordinates of the original image (Analyze
format) and -oc are the voxel coordinates (in coordinate space of
original image) or the desired origin, e.g. the AC. This can be obtained
via
display in SPM. Note, the .hdr file is not properly read, so
modifying the original SPM header doesn't help anything. Figure out
what the new voxel origin is, so that the correct values can be given
to mri_convert.
This step is necessary for proper display of surfaces if the origin in
the original GE volume is fairly far from AC.
Translations for modfmri subjects:
28jan02MC: -ic 62 128 128 -oc 61 167 144
28jan02PJ: -ic 62 128 128 -oc 65 148 134
*NOTE* I discovered that this conversion does a L/R flip when going
from the Analyze format file generated by the original GE2SPM
conversion program. To prevent the flip, one has to set the RAS
directions on the input using the -iid, -ijd, and -okd flags.
4. The COR volume in the orig directory is now used for FreeSurfer
processing. Note that the origin is a bit off of the AC, but this
isn't critical.
5. Create a new analyze format file which precisely matches the
FreeSurfer COR volume. This file will then be used for alignments of
functionals, etc. in SPM.
mri_convert /data2/modfmri/28jan02MC/mri/orig/
/data2/modfmri/28jan02MC/hires/28jan02MC_002.img
*** NOTE *** It turns out that mri_convert creates a screwed up
header. In order to fix things, a couple the .hdr and the .mat files
have to be fixed as follows:
% Read in the original .hdr
[DIM VOX SCALE TYPE OFFSET ORIGIN DESCRIP] =
spm_hread('28jan02MC_COR.hdr');
ORIGIN = [128 128 128];
% Write out the new header
spm_hwrite('28jan02MC_COR.hdr', DIM,VOX,SCALE,TYPE,OFFSET,ORIGIN,DESCRIP);
% Create a new matfile
M = [1.0000 0 0 -129.0000;
0 0.0000 1.0000 -129.0000;
0 -1.0000 0.0000 129.0000;
0 0 0 1.0000];
save('28jan02MC_COR', 'M')
Cheers,
Petr
On Wed, 1 May 2002, Frederick Klauschen wrote:
> I have now tried another brain which has a perfect
> wm segmentation. Also, "filled wm" looks nice.
> The create surface process does not produce any error
> message.
> But still, after clicking on SURFACE, only a black
> empty window appears.
>
> Frederick
>
>
> __________________________________________________
> Do You Yahoo!?
> Yahoo! Health - your guide to health and wellness
> http://health.yahoo.com
>
