External Email - Use Caution
Hey Sebastien,
I'm not sure how to do this in FreeSurfer (mri_add_xform_to_header might work)
but you can overwrite the header in python using nibable. You'll have to
compute the vox2ras yourself however.
Let's say you have image 1 and image 2 and you want to overwrite the header in
image 2. I would set the grad2ras matrix (what mri_info calls 'xform info') of
image 2 to be what is in image 1, then use the voxel sizes of image 2 to
compute the vox2ras from the grad2ras.
Here is some code (from the autoprescribe notebook!) showing how to go from
vox2ras to grad2ras, you'll need to go in the opposite direction (grad2ras -->
vox2ras)
```
# Given
# - A nibable img as returned by nb.load()
# Compute
# - The vox2ras matrix (this is simply img.affine)
# - The grad2ras matrix (mri_info calls this xform info)
# - See mri.cpp:12818
def calc_transforms(img):
vox2ras = np.copy(img.affine)
grad2ras = np.copy(vox2ras)
# - pixdims is same as xsize, ysize, zsize in mri.cpp:12818
# - c is same as c_r, c_a, c_s in mri.cpp:12818
# - use of 1:4 here is probably fragile
pixdims = img.header['pixdim'][1:4]
c = img.header['dim'][1:4] / 2
grad2ras[0:3,0:3] = vox2ras[0:3,0:3] / pixdims
grad2ras[0:3,3] = vox2ras[0:3,3] + vox2ras[0:3,0:3] @ c
return vox2ras, grad2ras
```
Good luck,
-Paul
________________________________________
From: freesurfer-boun...@nmr.mgh.harvard.edu
<freesurfer-boun...@nmr.mgh.harvard.edu> on behalf of Proulx, Jean Sebastien
<jspro...@mgh.harvard.edu>
Sent: Friday, September 8, 2023 9:10 AM
To: freesurfer@nmr.mgh.harvard.edu
Subject: Re: [Freesurfer] Copy geometry (fov position) from one header to
another while keeping voxel and matrix sizes unchanged
External Email - Use Caution
Dear Freesurfers,
Any chance anyone has a suggestion to achieve that? I’m a bit stuck.
Summary: two runs of a single-slice image don’t aligned in RAS space, but they
do in brain space (at the scanner, the head moved between the runs but the FOV
was adjusted accordingly with autoalign-like routine). The two runs have
different voxel and matrix sizes. I want to copy the geometry information from
one run to the other without resampling the data, so that the two can be
properly overlayed in freeview. mri_convert with the --in_like flag does not
work with unmatched voxel grids.
Thanks for you help and
Have a great day!
Sebastien
From: Proulx, Jean Sebastien <jspro...@mgh.harvard.edu>
Date: Monday, August 28, 2023 at 3:12 PM
To: freesurfer@nmr.mgh.harvard.edu <freesurfer@nmr.mgh.harvard.edu>
Subject: RE: Copy geometry (fov position) from one header to another while
keeping voxel and matrix sizes unchanged
Exactly, the two runs align in voxel space but not in RAS space.
mri_convert with the --in_like did the trick, thank you very much.
Now I have another scenario where the two runs also don’t align in RAS space
but where they would align in voxel space if they had the same voxel grid
dimensions (the second run has twice the spatial resolution). I need to achieve
the same thing: adjust geometry in the header of the second run—without
changing the data—so it aligns in RAS space with the first run.
mri_convert with --in_like unfortunately complains for that second scenario:
“error: volume sizes do not match”
Thanks again a lot for your help and
Have a great day!
Sébastien
From: Greve, Douglas N.,Ph.D. <dgr...@mgh.harvard.edu>
Sent: Monday, August 28, 2023 10:47 AM
To: Proulx, Jean Sebastien <jspro...@mgh.harvard.edu>;
freesurfer@nmr.mgh.harvard.edu
Subject: Re: Copy geometry (fov position) from one header to another while
keeping voxel and matrix sizes unchanged
To confirm, the two volumes *without any transformation (ie, "untouched")* do
not align in FV but they do align in fslview? In other words, they align in
voxel space but not in RAS space?
To just copy the geometry, run
mri_convert run2.nii.gz --in_like run1.nii.gz run2-in-run1.nii.gz
On 8/28/2023 10:10 AM, Proulx, Jean Sebastien wrote:
Sorry for sending to the wrong place, will do in the future. Here it is.
Thanks for the suggestion! The two fov indeed don't align in freeview. I played
a little with the --no-resample option but did not acheive what I wanted.
Here I simplified the problem to two single-slice images (run1.nii.gz and
run2.nii.gz attached) acquired the same way (same resolution) but in two
different positions in scanner space. In freeview it looks like this:
freeview anat.nii.gz run1.nii.gz run4.nii.gz
[Image]
I need the fov to align, but the data to not change at all (because the fov
moved with the patient's head during the experiment).
To get the transformation, I ran
tkregister2 --mov run4.nii.gz --targ run1.nii.gz --regheader --reg run4to1
and clicked "save reg" without touching anything. It spits out the following
matrix
0.99988 -0.01117 0.01053 -0.63259;
0.01065 0.99879 0.04792 -0.70029;
-0.01105 -0.04780 0.99880 -3.93196;
0.00000 0.00000 0.00000 1.00000;
Is this indeed the run4 to run1 transformation matrix? Do you know of a way to
get that transformation without user information?
Then I thought
mri_vol2vol --mov run4.nii.gz --targ run1.nii.gz --reg run4to1 --no-resample
--o run4to1.nii.gz
would do the trick but no, the run1.nii.gz and run4to1.nii.gz don't align, just
like in the image above, as if the transformation is not actually applied.
Allowing resample
mri_vol2vol --mov run4.nii.gz --targ run1.nii.gz --reg run4to1 --o
run4to1.nii.gz
makes the two field of view align in freeview but the data is altered.
[Image]
It looks like the transformation is not actually applied so that only the
intersecting voxels are effectively resampled.
There is no error message.
I tried the dev environment and stable711.
I tried renaming the run4to1 registration file to run4to1.lta and I get the
folloing error from mri_vol2vol:
regio_read_register: loading lta
error: Cannot allocate memory
error: LTAalloc(675792816): could not allocate xforms
I tried renaming the run4to1 registration file to run4to1.xfm and nothing
changes (no error but the registration still does not seem to be applied).
Any idea what's going on?
Thanks a lot for your help and
Good evening!
Sebastien
Douglas N. Greve Mon, 28 Aug 2023 06:52:20 -0700
External Email - Use Caution Sorry, I don't have your previous email so I don't
know what the context is. Can you resend this email with the previous exchanges
included at the bottom? Also, please remember to send to the FS list instead of
us personally.
thanks!
doug
Douglas N. Greve Sun, 20 Aug 2023 14:52:01 -0700
External Email - Use Caution Do the two align in freeview if you just load both
of them? If so, then you can use something like mri_vol2vol --mov vol1.mgz
--targ vol2.mgz --regheader --no-resample --o new.vol1.mgz If they do not but
you have a registration, then you can use a similar command, eg, mri_vol2vol
--mov vol1.mgz --targ vol2.mgz --reg yourreg.lta --no-resample --o new.vol1.mgz
On 8/18/2023 11:34 AM, Proulx, Jean Sebastien wrote:
External Email - Use Caution
Dear Freesurfers!
I have two single-slice images that were acquired with different head positions.
Thanks to a fancy AutoAlign-like routine at the scanner (areg by Paul Wighton),
the fov of the second image tracked the movement of the head—the brain nicely
aligns when overlaying the two untouched image grids with e.g. fslview.
To leverage this online registration and properly visualize images in freeview,
I need the fov position in header of the second image to be the same as in the
header of the first image—basically pretending the head did not move and the
two images were acquired with the exact same scanner-space fov position. So far
I achieved that in a somewhat inelegant way by loading the first header in
matlab with MRIread.m and manually adding the volume data from the second image
before saving with MRIwrite.m.
That however does not work when the second image have different voxel and
matrix sizes—e.g. my second image has twice the resolution of the first and
gets blown up to twice the size in freeview after my makeshift header change :-(
I suspect a better solution could use mri_vol2vol with an appropriate
registration matrix, but I don't know how to get the appropriate matrix.
Any suggestions?
thanks a lot for the help and
Have a very good day!
Sébastien
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Mass General Brigham
Compliance HelpLine at https://www.massgeneralbrigham.org/complianceline
<https://www.massgeneralbrigham.org/complianceline> .
Please note that this e-mail is not secure (encrypted). If you do not wish to
continue communication over unencrypted e-mail, please notify the sender of
this message immediately. Continuing to send or respond to e-mail after
receiving this message means you understand and accept this risk and wish to
continue to communicate over unencrypted e-mail.
