[FieldTrip] Fwd: MOUS dataset source and head models
Schoffelen, J.M. (Jan Mathijs)
janmathijs.schoffelen at donders.ru.nl
Mon Jan 3 11:11:15 CET 2022
Hi Philip,
When thinking about surface meshes, there are 2 things to take into account:
1) the positions of the vertices, and
2) the way the vertices are connected (i.e. the edges), which is typically defined in terms of the triangles.
As long as the set of vertices is well-defined, i.e. each one of them is defined in relation to the other one (e.g. by giving them a unique index number, ranging from 1 to, say, 8196), the absolute position of the vertices can be played with. Also, importantly, as long as a 8196-node cortical surface is topologically equivalent to another 8196-node (note that this is not a given) surface, a parcellation scheme (e.g. atlas) defined on one surface can be directly applied to the other one (see below).
Of course, in real life situations, the positions of the set of vertices are defined in some coordinate system that relates to a realistic anatomical shape, but for the template registration this is actually irrelevant (and nor is it for the visualization). For this reason, often surface-based results are displayed on an ‘inflated’ cortical surface, to allow for looking into the sulci. Thus, the - what you refer to as template - cortical surface that you attached to your previous message is an inflated cortical surface, that is by definition not defined in any real-world coordinate system.
Surface based template matching takes place in topological space, where the subjects’ hemisphere specific meshes, which are topological spheres, are matched to a spherical template. This spherical template has the most important cortical landmarks defined, to which the individual surfaces are morphed. These landmarks are fissures/sulci etc. This registration procedure is fundamentally different from a volume based registration procedure.
Now, I am not 100% sure that the 8196 node template in the fieldtrip distribution is in the same topological space as the topological space used for the MOUS subjects, but it could be the case (I am pretty sure it is). The reason I mention it, is that you give the impression that you’d like to define ROIs in MNI space, and use the fieldtrip template to identify the relevant vertex indices, in order to transport those indices to the data from MOUS. In general, I would discourage people to combine surface-based source models with ROIs (or atlases) defined in volumetric space, but it’s of course not forbidden. Should you want to walk that path, you indeed need to ensure that the vertex positions of template surface are in a well-defined coordinate space.
To verify this, you may need to visualize the cortical mesh in an overlay on top of an anatomical image that is in the desired coordinate space.
I quickly checked to verify that the surface mesh in FT surf = ft_read_headshape('cortex_8196.surf.gii’);
is in register with the anatomical template: mri = ft_read_mri(’single_subj_T1_1mm.nii’);
figure;ft_plot_ortho(mri.anatomy,’transform’,mri.transform,’intersectmesh’,surf);
Provided the template anatomy is in standardized space, and the surface mesh is in topological register with the ones used for MOUS, you should be good to go.
Best wishes,
Jan-Mathijs
[cid:D04A8CC9-5D25-4AFC-8849-1225780A5C90]
,
On 2 Jan 2022, at 08:50, Philip Cho <philip.cho at duke.edu<mailto:philip.cho at duke.edu>> wrote:
The question I just asked with respect to the warping is of interest to me, but I'm now putting together that it's not very relevant. The part that I had missed was that the dipole positions for individual-subject meshes are already comparable across subjects, since they supposedly are surface registered to a common template. That being said, I'm slightly confused now as to which template this is. The dipole position indices from the individual meshes (provided from your group) do not seem to correspond to the 8196-vertex template provided with FieldTrip, but instead, they correspond to the indices in "cortex_inflated_8196reg.mat", which is provided with the dataset. Is this correct? The template from FieldTrip is expressed in MNI coordinates, so it was easy to apply ROIs and such, but I'm noticing that the other template (supposedly the correct one) is more ambiguous, given that it's shifted to the left, as depicted below. If it could be clarified if this is the correct template, and if so, what coordinate system it's in, I think this would be the last bit of help I would need. Thank you so much so far! My understanding has improved tremendously.
<image.png>
________________________________
From: Philip Cho <philip.cho at duke.edu<mailto:philip.cho at duke.edu>>
Sent: Sunday, January 2, 2022 1:28 AM
To: Schoffelen, J.M. (Jan Mathijs) <janmathijs.schoffelen at donders.ru.nl<mailto:janmathijs.schoffelen at donders.ru.nl>>
Subject: Re: MOUS dataset source and head models
Hi Jan-Mathijs,
Thanks so much for your response. I am subscribed to the discussion list and regularly receive email notifications, but I was never able to see your original response, and I still cannot seem to find it. Regardless, I really apologize that you had to answer my question twice. With regard to your response, I understand that the warped template won't exactly match the individual-subject meshes, but I thought the general size/outline of the two meshes would be more similar, given that the template was warped based on the individual subject's MRI. The warping seems to orient the brain correctly, but the shape/size of the mesh remains vastly different, which makes me think the warping did not work completely as intended. Please let me know if there's something I'm still not understanding here. Thank you so much!
Best,
Philip
________________________________
From: Schoffelen, J.M. (Jan Mathijs) <janmathijs.schoffelen at donders.ru.nl<mailto:janmathijs.schoffelen at donders.ru.nl>>
Sent: Tuesday, December 28, 2021 3:59 AM
To: Philip Cho <philip.cho at duke.edu<mailto:philip.cho at duke.edu>>
Subject: Re: MOUS dataset source and head models
Hi Philip,
If I recall well, I have already answered to the e-mail you copied in below. I forwarded it to the discussion list, and answered it there. If you haven’t received that answer, then I strongly suggest you to subscribe to the discussion list. If you received the answere, then I don’t know what there is to add. In that case you need to formulate your question in a different way, because then I don’t understand what the problem is.
Best wishes,
Jan-Mathijs
Hi Philip,
I think it makes sense to discuss this issue on the mailing list.
The difference that you describe arise from the fact that the subject individual headmodels are NOT a spatially warped version of the template grid. The individual models were constructed from the individual subjects’ anatomical images. If it is the intention to perform a source analysis on the cortical surface level, it’s always preferred (if possible) to use the subject-specific meshes, and not a spatially warped (i.e. warped by means of spatial transformations that are applicable to 3D volumetric objects) template. This is because the gyrification is of course highly individual.
Best wishes,
Jan-Mathijd
On 28 Dec 2021, at 08:32, Philip Cho <philip.cho at duke.edu<mailto:philip.cho at duke.edu>> wrote:
Hi Jan-Mathijs,
I apologize if this is not the most appropriate way to contact you, but I've worked at this issue for some time now and still don't quite understand what the issue is. It's quite critical that I get this part right, so I hope it's understandable that I'm reaching out for your input. Any help would be appreciated. Thank you so much!
Best,
Philip
________________________________
From: Philip Cho
Sent: Tuesday, November 9, 2021 4:20 PM
To: jan.schoffelen at donders.ru.nl<mailto:jan.schoffelen at donders.ru.nl> <jan.schoffelen at donders.ru.nl<mailto:jan.schoffelen at donders.ru.nl>>
Subject: MOUS dataset source and head models
Hi Jan-Mathijs,
I hope you're doing well. I once emailed you a long time ago about obtaining source and head models for LCMV on the MOUS dataset, and you kindly provided them. I've done a lot of work with source-level analysis since then, and the approach I've generally used is to use the 8196-vertex template mesh defined in MNI space and warping it to subject space using the subject's MRI. I was using this approach in order to average these results across subjects. However, recently I have found that the warped meshes look slightly different from the source models that you have provided. I assumed that they would be more spatially similar. In particular, the warped mesh seems to be more constricted (smaller) and pushed slightly farther back. In comparison, the source model that was provided has points that push farther forward towards the front of the head volume. I'm curious why these differences arise? I have looked into it for some time now with little success, and I think it would help me going forward to understand this. I have attached my warping code below.
Thank you so much!
Best,
Philip
cd(['D:\MOUS\data\sub-' sub{1} '\anat']);
load([sub{1} 'vol.mat']);
mri = ft_read_mri(['sub-' sub{1} '_space-CTF_T1w.nii']);
mri.coordsys = 'ctf';
% create the subject specific grid, using the template grid that has just been created
cfg = [];
cfg.method = 'basedonmni';
cfg.template = template_grid;
cfg.nonlinear = 'yes';
cfg.mri = mri;
cfg.unit ='mm';
sourcemodel_warpedTemplate = ft_prepare_sourcemodel(cfg);
save(['sourcemodel_warpedTemplate_surface8196_' sub{1} '.mat'],'sourcemodel_warpedTemplate');
% make a figure of the single subject headmodel, and grid positions
figure; hold on;
ft_plot_headmodel(vol, 'edgecolor', 'brain', 'facealpha', 0.4);
ft_plot_mesh(sourcemodel_warpedTemplate.pos(sourcemodel_warpedTemplate.inside,:));
savefig(gcf,['warpedsourceSurface8196_head_overlay_' sub{1} '.fig']);
close all
<image.png><image.png>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20220103/80dff50d/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Screenshot 2022-01-03 at 10.49.13.png
Type: image/png
Size: 918914 bytes
Desc: Screenshot 2022-01-03 at 10.49.13.png
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20220103/80dff50d/attachment-0001.png>
More information about the fieldtrip
mailing list