Hi Guillaume,
Thank you very much for your suggestion and sorry for the delayed response - I was trying to learn as much as I could about using SPM12 before posting any other question. And yes the spm_swarp.m function that you pointed me to looks like exactly what I need: that is to use the deformation field to transform native coordinates to MNI space. But I must be doing something wrong because the output mesh that spm_swarp returns has about half the vertices equal to 0.
Looking at my processing below, would you have any idea what the problem could be? Also would greatly appreciate any example of correctly using spm_swarp or any other function that uses the deformation field to generate a warped mesh.
Here's what I do (again I only deal with structural NIFTI volumes, no fMRI's).
T1.nii --> Segmentation --> c1T1.nii
T1.nii, c1T1.nii --> Coregistration --> rT1.nii, rc1T1.nii
rT1.nii, rc1T1.nii --> Normalization --> wrT1.nii, wrc1T1.nii, y_rT1.nii
After this processing, I want to apply the deformation field y_rT1.nii to the subject's brain mesh to get the corresponding MNI coordinates.
This is the code I use:
%%%%% Finished running Segmentation, Coregistration, Normalization. Now apply the
%%%%% deformation field to the coregistered brain mesh
% First generate brain mesh from coregistered structural brain volume
brain_nii = MRIread([subj_folder, '/rc1T1.nii']);
brain_mesh = isosurface(brain_nii.vol, 0.9);
% isosurface flips x and y, so flip it back to be consistent with voxel space
brain_mesh.vertices = [brain_mesh.vertices(:,2), brain_mesh.vertices(:,1), brain_mesh.vertices(:,3)];
% Generate warped mesh using spm_swarp. Use similar code as spm_deformations.m
% which also calls spm_swarp during normalization. Copied to here, from
% spm_deformations.m, the internal functions:
%
% get_comp, get_job, get_def, get_idbbvox, get_idbbvox, and identity
%
% so we can mimic what spm_deformations does
job{1}.def = {[subj_folder, '/y_rT1.nii']};
job{2}.idbbvox = matlabbatch{3}.spm.spatial.normalise.estwrite.woptions;
[def, T] = get_comp(job);
wbrain_mesh = spm_swarp(brain_mesh, def, T);
% convert gifti to struct
wbrain_mesh_s.vertices = wbrain_mesh.vertices;
wbrain_mesh_s.faces = wbrain_mesh.faces;
When this is done, I see that the number of vertices and faces in the original and warped meshes matches. But when I view random vertices in the warped mesh I see row after row of [0,0,0]. Also when I view the original mesh against the warped mesh in a figure, I see a very strange looking brain - only about a quarter to an 8th piece of the warped brain.
Any help would be very appreciated. For reference I include below, the job parameters I used to create the output.
Jay
====================================================
spm('defaults','fmri');
spm_jobman('initcfg');
clear matlabbatch
%%%%% Segmentation
matlabbatch{1}.spm.spatial.preproc.channel.vols = {[subj_folder, '/T1.nii,1']};
matlabbatch{1}.spm.spatial.preproc.channel.biasreg = 0.001;
matlabbatch{1}.spm.spatial.preproc.channel.biasfwhm = 60;
matlabbatch{1}.spm.spatial.preproc.channel.write = [0 0];
matlabbatch{1}.spm.spatial.preproc.tissue(1).tpm = {'c:\spm12\tpm\TPM.nii,1'};
matlabbatch{1}.spm.spatial.preproc.tissue(1).ngaus = 1;
matlabbatch{1}.spm.spatial.preproc.tissue(1).native = [1 0];
% run segmentation separately rather than as part of one big job - for debugging purposes
spm_jobman('run',matlabbatch(1));
%%%%% Coregistration
matlabbatch{2}.spm.spatial.coreg.estwrite.ref = {[mni_folder, '/mni152/T1.nii,1']};
matlabbatch{2}.spm.spatial.coreg.estwrite.source = {[subj_folder, '/T1.nii,1']};
matlabbatch{2}.spm.spatial.coreg.estwrite.other = {
[subj_folder, '/c1T1.nii,1']
[subj_folder, '/c2T1.nii,1']
[subj_folder, '/c3T1.nii,1']
};
matlabbatch{2}.spm.spatial.coreg.estwrite.eoptions.cost_fun = 'nmi';
matlabbatch{2}.spm.spatial.coreg.estwrite.eoptions.sep = [4 2];
matlabbatch{2}.spm.spatial.coreg.estwrite.eoptions.tol = [0.02 0.02 0.02 0.001 0.001 0.001 0.01 0.01 0.01 0.001 0.001 0.001];
matlabbatch{2}.spm.spatial.coreg.estwrite.eoptions.fwhm = [7 7];
matlabbatch{2}.spm.spatial.coreg.estwrite.roptions.interp = 4;
matlabbatch{2}.spm.spatial.coreg.estwrite.roptions.wrap = [0 0 0];
matlabbatch{2}.spm.spatial.coreg.estwrite.roptions.mask = 0;
matlabbatch{2}.spm.spatial.coreg.estwrite.roptions.prefix = 'r';
% run coregistration separately rather as part of one big job - for debugging purposes
spm_jobman('run',matlabbatch(2));
%%%%% Normalization
matlabbatch{3}.spm.spatial.normalise.estwrite.subj.vol = {[subj_folder, '/rT1.nii,1']};
matlabbatch{3}.spm.spatial.normalise.estwrite.subj.resample = {
[subj_folder, '/rT1.nii,1']
[subj_folder, '/rc1T1.nii,1']
};
matlabbatch{3}.spm.spatial.normalise.estwrite.eoptions.biasreg = 0.0001;
matlabbatch{3}.spm.spatial.normalise.estwrite.eoptions.biasfwhm = 60;
matlabbatch{3}.spm.spatial.normalise.estwrite.eoptions.tpm = {'c:\spm12\tpm\TPM.nii']};
matlabbatch{3}.spm.spatial.normalise.estwrite.eoptions.affreg = 'mni';
matlabbatch{3}.spm.spatial.normalise.estwrite.eoptions.reg = [0 0.001 0.5 0.05 0.2];
matlabbatch{3}.spm.spatial.normalise.estwrite.eoptions.fwhm = 0;
matlabbatch{3}.spm.spatial.normalise.estwrite.eoptions.samp = 3;
matlabbatch{3}.spm.spatial.normalise.estwrite.woptions.bb = [-127 -127 -127
128 128 128];
matlabbatch{3}.spm.spatial.normalise.estwrite.woptions.vox = [1 1 1];
matlabbatch{3}.spm.spatial.normalise.estwrite.woptions.interp = 4;
matlabbatch{3}.spm.spatial.normalise.estwrite.woptions.prefix = 'w';
% run normalization separately as part of one big job - for debugging purposes
spm_jobman('run',matlabbatch(3));
________________________________
From: Guillaume Flandin <[log in to unmask]>
Sent: Wednesday, August 1, 2018 5:20:02 AM
To: Dubb, Jay
Cc: [log in to unmask]
Subject: Re: [SPM] Advice for SPM beginner on how to normalize structural volumes
Dear Jay,
Have a look at spm_mesh_transform.m (a basic wrapper above spm_swarp.m);
the coordinates will have to be stored in a 'vertices' field of a struct.
Best regards,
Guillaume.
On 31/07/18 20:26, Dubb, Jay wrote:
>
> Hi Tibor,
>
>
> Thanks for the links and tips. Now that I've read up on these topics and
> played around a bit more with SPM, I have a more specific question:
>
>
> I ran Normalization in SPM12 for my structural volume named rT1.nii and
> got as output the files y_rT1.nii and wrT1.nii. I checked the
> registration of the normalized subject volume (wrT1.nii) against the
> MNI152 template using the SPM Check Reg and see that it looks accurate.
>
>
> My question is, do you know how I would apply the output of
> normalization in a Matlab script to the coordinates of the initial
> subject volume rT1.nii, to calculate the corresponding MNI coordinates.
>
>
> Thanks again,
>
> Jay
>
>
>
>
>
>
> ------------------------------------------------------------------------
> *From:* SPM (Statistical Parametric Mapping) <[log in to unmask]> on
> behalf of Auer, Tibor <[log in to unmask]>
> *Sent:* Tuesday, July 31, 2018 7:14 AM
> *To:* [log in to unmask]
> *Subject:* Re: [SPM] Advice for SPM beginner on how to normalize
> structural volumes
>
>
> Hi Jay,
>
>
>
> Ad a.: I recommend you to read the corresponding section of the SPM
> manual and/or the SPM book
> (https://www.fil.ion.ucl.ac.uk/spm/doc/books/hbf2/). There also other
HUMAN BRAIN FUNCTION 2nd EDITION - fil.ion.ucl.ac.uk<https://www.fil.ion.ucl.ac.uk/spm/doc/books/hbf2/>
www.fil.ion.ucl.ac.uk
Human Brain Function, 2nd edition. 2 nd EDITION. Section Editors. J Ashburner K Friston W Penny. PART II - IMAGING NEUROSCIENCE - THEORY AND ANALYSIS
> useful online materials. Se e.g. mine:
> http://www.cubic.rhul.ac.uk/wiki/lib/exe/fetch.php?media=fmri:5_pre-processing.pdf(slide
> 9 and onwards).
>
>
>
> Ad b.: You can try mailing list archive. E.g. for DARTEL (you mentioned
> non-linear transformations):
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=spm;2d6e16b4.1202
>
>
>
> Vale,
>
> Tibor
>
>
>
> *Auer, Tibor M.D. Ph.D.*
>
> *Research Fellow in MRI*
>
> *Department of Psychology*
>
> Royal Holloway University of London
>
> Egham Hill
>
> Egham
>
> TW20 0EX
>
> United Kingdom
>
> +44-1784-27-*6594*
>
> [log in to unmask] <mailto:[log in to unmask]>
>
>
>
>
>
> *From:*SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
> *On Behalf Of *Dubb, Jay
> *Sent:* 30 July 2018 23:39
> *To:* [log in to unmask]
> *Subject:* [SPM] Advice for SPM beginner on how to normalize structural
> volumes
>
>
>
>
>
> Hello dear SPM users,
>
>
>
> I am new to SPM. I am looking for advice about some relatively simple
> examples of how one could use the SPM12 batch script to convert
> structural volume native coordinates, say T1.nii, to MNI space. My
> MATLAB graphical application displays individual subjects' head
> and brain in 3D in their native space. It allows users to hide the head
> and click on any location on the brain. I want to add the ability to
> display the MNI152 coordinates which correspond to the selected
> native cortical locations. (Even though it's a MATLAB application, it
> has some parts which tie it to Windows, therefore I am not looking at
> applications such as FSL and FNIRT, which are Linux-based.)
>
>
>
> I was told that I have to do segmentation and then normalization. Maybe
> coregistration as well, although unclear about that too. Or can I do
> ONLY the normalization step with T1.nii as input?
>
>
>
> My questions are:
>
>
>
> a) What steps are needed to do the MNI normalization for my purposes (I
> only care about single subject structural NIFTI volumes, No fMRI)
>
>
>
> b) How would one use the output of the normalization workflow in a batch
> script to do the coordinate transformation to MNI? Since normalization
> involves warping, and non-linear transformations, it's not as simple as
> applying an affine 4x4 transformation matrix to my native coordinates
> ... as far as I understand.
>
>
>
> Any advice would be greatly appreciated...
>
>
>
> Jay
>
>
>
>
>
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
