The correspondence of vertex outside medial wall between Yeo atlas and fs_LR 32k template

[Zack Guo]

Dear HCP experts,

I'm confused about the correspondence of vertices outside the medial wall between the Yeo atlas on fs_LR 32k and standard fs_LR 32k surface.

Here are the files I used:
Yeo atlas file: https://balsa.wustl.edu/study/show/WG33 (file name: RSN-networks.32k_fs_LR.dlabel.nii)
Standard fs_LR 32k template files: https://github.com/Washington-University/HCPpipelines/tree/master/global/templates/91282_Greyordinates (file name: {L, R}.atlasroi.32k_fs_LR.shape.gii)

Taking the left cortex as an example, some vertices (e.g. 163~166) in the standard template outside the medial wall correspond to the medial wall in the Yeo atlas; conversely, some vertices (e.g. 22~29) in the Yeo atlas outside the medial wall correspond to the medial wall of the standard template.

My goal is to find the correspondence non-medial wall vertices in standard templates and Yeo atlas. How should I solve the above problem? 

Thanks.

Best,

Zack

[Tim Coalson]

The RSN-networks dlabel file has not excluded the medial wall at the cifti xml level, so the cifti data contains all the vertices (the overlaid medail wall appears to be a separate file, and some of the label maps have their own label for the medial wall, which are different).  You can extract the ROI of the medial wall label with wb_command -cifti-label-to-roi using the extra options "-name FreeSurfer_Defined_Medial_Wall -map 2", and use -cifti-separate to convert that to gifti if desired.

If you just want it to match index-to-index with 91k cifti data, you can use wb_command -cifti-create-dense-from-template with the -cifti option, and the pipelines 91k cifti file as the template.

Tim

--
You received this message because you are subscribed to the Google Groups "HCP-Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hcp-users+...@humanconnectome.org.
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/754730a3-e46c-4393-a180-daa7e7c686d8n%40humanconnectome.org.

[Zack Guo]

Hi Tim,

Thank you for your reply. Yeah, I am interested in matching vertices in Yeo atlas to fs_LR 32k template. The command wb_command -cifti-create-dense-from-template is helpful. Some issues:  

1. I don't fully understand the matching process, how are the vertices of the Yeo atlas processed?
2. For Yeo atlas and fs_LR 32k template, the number of vertices outside the medial wall is not exactly the same (59412 and 50330). How to understanding this?

Best,

Zack

[Tim Coalson]

1. I don't fully understand this question, what do you mean by "processed"?  The version in balsa is already on the same 32k surface we use for nearly all of our cifti data (though the medial wall shift compared to the standard 91k masks suggests that the registration/resampling may not be entirely optimal).

2. How did you derive the 50330 number?  I get only 746 medial vertices (sum of left + right) outside the 91k standard medial wall from the 17 networks map, which means there should be 58666 91k-included vertices outside the Yeo medial wall.

Tim

[Zack Guo]

Hi Tim, 

Thanks for your detailed reply sincerely.

Sorry, my description was not clear enough. After reading your reply carefully and checking my results, I re-describe my confusion.

Before matching, the number of vertices of the Yeo atlas was 64984, and after matching it was 58666 (with network labels), a reduction of 6318. I want to figure out this change. My understanding is that the 5572 medial wall vertcies defined by standard 91k (64984-59412=5572) are simply discarded. And I find that 746 of the remaining 59412 vertices have  value of 37 (FreeSurfer_Defined_Medial_Wall). Is this due to registration/sample you mentioned before, or some other reason?

At the same time, I found another version of Yeo 7/17 atlas with some differences(https://github.com/ThomasYeoLab/CBIG/tree/master/stable_projects/brain_parcellation/Yeo2011_fcMRI_clustering/1000subjects_reference/Yeo_JNeurophysiol11_SplitLabels/fs_LR32k), file name: Yeo2011_{7, 17}Networks_N1000.dlabel.nii. Which version is more recommended?

Best, 

Zack

[Tim Coalson]

Yes, those 746 are because the medial wall definitions aren't the same, and to me, it looks mostly like a shift.  I don't know exactly what led to this difference.

For cortical data, we usually recommend our HCP MMP 1.0, not Yeo.  I don't know the conversion specifics of either version of Yeo.

Tim

[Zack Guo]

Hi Tim,

Thank you.

1. Yes, it looks like a shift of medial wall from a visualization view. 

2. Thanks again.

Best, 

Zack