ROI-ROI functional connectivity analyses

384 views
Skip to first unread message

Panos Fotiadis

unread,
Oct 7, 2021, 3:52:17 AM10/7/21
to HCP-Users
Hello,

I'm interested in generating ROI-ROI resting-state functional connectivity matrices using a Schaefer atlas and the Glasser multi-modal atlas, and I had a couple questions concerning the wb_command -cifti-parcellate command: 

1. I first demeaned and normalized the 4 individual *_Atlas_MSMAll_hp2000_clean.dtseries.nii files for each subject and then concatenated them into one file, as nicely described in the HCP Users FAQ. Does using the merged time series file as my input for the wb_command -cifti-parcellate command sound correct?

2. Concerning the parcellations that I'm interested in using, I wanted to verify that I'm feeding the appropriate dlabel files into my cifti-parcellate command: Are the Schaefer dlabels kindly provided here: https://github.com/ThomasYeoLab/CBIG/tree/master/stable_projects/brain_parcellation/Schaefer2018_LocalGlobal/Parcellations/HCP and the Glasser Parcellation: Q1-Q6_RelatedValidation210.CorticalAreas_dil_Final_Final_Areas_Group_Colors.32k_fs_LR.dlabel.nii the most appropriate files to use as dlabels into my cifti-parcellate command? I was a little confused since my input file contains 91282 vertices, whereas the Schaefer and Glasser files contain 64984 and 59412 vertices, respectively. Would the command take care of this mismatch in both cases (which from what I've read is due to the presence/absence of the medial wall?) or would I need to come up with dlabel files that have the same number of vertices as my input file?  

Thank you in advance for your time!

Cheers,
Panos

Glasser, Matt

unread,
Oct 7, 2021, 5:50:50 AM10/7/21
to hcp-...@humanconnectome.org
  1. That should be fine.
  2. I can’t help with the Schaefer parcellation as I don’t know the details of that.  The HCP’s multi-modal parcellation is cortex only; however, we have sometimes added the standard FreeSurfer subcortical structures to it.  If that is of interest (this is 91282), I have uploaded it to BALSA: https://balsa.wustl.edu/file/87B9N else the cortex only version is here: https://balsa.wustl.edu/file/3VLx

 

Matt.

--
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/5be40db1-985f-4f29-b503-9bc29fdf566bn%40humanconnectome.org.

 

The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.

Panos Fotiadis

unread,
Oct 7, 2021, 12:54:37 PM10/7/21
to HCP-Users, glas...@wustl.edu

Awesome, many thanks Matt! To follow up:

1. So both Glasser parcellations you provided should be fine being fed to my wb_command -cifti-parcellate as is, even though the cortical version has a different set of vertices than my input file, right? (the command itself runs with no error - I just wanted to make sure that the results were sensical.)

2. This is semi-relevant but: What is the recommended way to transfer the Glasser parcellations to an individual subject's orig space (for instance, something like the equivalent of aparc+aseg.mgz)? I would assume that mri_surf2surf followed by mri_aparc2aseg would be the way to go, but I wasn't sure which glasser file to use as input (or whether there is a more appropriate approach).

Thanks again,
Panos

Joseph Orr

unread,
Oct 7, 2021, 1:31:08 PM10/7/21
to HCP Users
This atlas has a subcortical hierarchical parcellation along with the Glasser, Schaefer, and Gordon cortical parcellations in CIFTI space: https://github.com/yetianmed/subcortex


--
Joseph M. Orr, Ph.D.
Assistant Professor
Department of Psychological and Brain Sciences


On Thu, Oct 7, 2021 at 11:54 AM Panos Fotiadis <pfo...@gmail.com> wrote:
Awesome, many thanks Matt! To follow up: 1. So both Glasser parcellations you provided should be fine being fed to my wb_command -cifti-parcellate as is, even though the cortical version has a different set of vertices than my input file, right? ZjQcmQRYFpfptBannerStart
This Message Is From an External Sender
This message came from outside your organization.
ZjQcmQRYFpfptBannerEnd

Glasser, Matt

unread,
Oct 7, 2021, 2:22:31 PM10/7/21
to Panos Fotiadis, HCP-Users
  1. Yes you will get a ptseries with only cortical results if you give a cortical parcellation.
  2. I don’t know what you mean by “orig space.” 

Glasser, Matt

unread,
Oct 7, 2021, 2:23:46 PM10/7/21
to hcp-...@humanconnectome.org

There are a variety of options out there.  I personally do not endorse any of them as better than any others of them.  I am not aware of a multi-modal subcortical parcellation made using a similar approach to the HCP’s cortical parcellation.


Matt.

Panos Fotiadis

unread,
Oct 7, 2021, 2:57:04 PM10/7/21
to HCP-Users, glas...@wustl.edu
Thank you both, Joseph and Matt!

Matt - By orig space I mean the individual subject's space as defined by Freesurfer (basically how after you run Freesurfer's recon-all on your subject, all Freesurfer-derived outputs including the default desikan-killiany parcellation files are on the same space as orig.mgz). So I was wondering whether there was a recommended approach to register the Glasser atlas from fs_LR space to an individual subject's space (as defined by the Freesurfer recon-all derived files under ${subject_folder}/${subject}/T1w/${subject}/).

Thanks again,
Panos

Glasser, Matt

unread,
Oct 7, 2021, 3:10:29 PM10/7/21
to hcp-...@humanconnectome.org

So you want to map the data into the individual’s volume space?  FreeSurfer is a bit weird about both coordinate spaces and their heavy reliance on the native mesh.  For the HCP we do things a bit differently and use a consistent coordinate space for all surfaces and volumes and do most of our analyses on standard meshes (which match the shape of the subjects surfaces but can be directly compared across subjects).  For volume spaces we have ${StudyFolder}/${Subject}/T1w which is rigidly aligned to MNI space (no change in brain size/shape, but a consistent origin and orientation) and nonlinearly aligned MNI space.  The native and 32k_fs_LR surfaces are available in both volume spaces and the 164k_fs_LR surfaces are in MNI space only.  For surface meshes we have FreeSurfer’s native mesh, and 164k and 32k standard meshes.  The 32k mesh is the main workhorse (2mm average vertex spacing). 

 

Matt.

 

From: Panos Fotiadis <pfo...@gmail.com>


Reply-To: "hcp-...@humanconnectome.org" <hcp-...@humanconnectome.org>
Date: Thursday, October 7, 2021 at 1:57 PM
To: HCP-Users <hcp-...@humanconnectome.org>
Cc: "Glasser, Matt" <glas...@wustl.edu>

Coalson, Timothy Scott (S&T-Student)

unread,
Oct 7, 2021, 5:19:57 PM10/7/21
to hcp-...@humanconnectome.org
You can use wb_command -cifti-separate on the dlabel to get gifti versions, then -label-to-volume-mapping to make left and right hemisphere label volumes (we recommend the ribbon constrained method with the individual's white and pial surfaces, in the space you want).  If you want to combine them, you will need to decide what to do if voxels overlap.

Note that this is only reasonable with individual surfaces, for use with that one individual's data.  Volume registration struggles in human cortex, and surface-based processing and registration avoids the 3D problem of aligning gray matter across subjects, and does better at aligning functional areas: https://www.pnas.org/content/115/27/E6356

I'm not sure how freesurfer's surface to volume methods work, but mri_surf2vol appears to depend on the voxel-based cortical mask, so I don't know if they use the pial and white surfaces to disambiguate.

Tim

From: Glasser, Matt <glas...@wustl.edu>
Sent: Thursday, October 7, 2021 2:10 PM
To: hcp-...@humanconnectome.org <hcp-...@humanconnectome.org>

Subject: Re: [hcp-users] ROI-ROI functional connectivity analyses
 
WARNING: This message has originated from an External Source. This may be a phishing expedition that can result in unauthorized access to our IT System. Please use proper judgment and caution when opening attachments, clicking links, or responding to this email.

Panos Fotiadis

unread,
Oct 9, 2021, 4:01:07 PM10/9/21
to HCP-Users, tsc...@mst.edu
That is extremely helpful, Matt and Tim, thank you! 

When I run the -label-to-volume-mapping command for each hemisphere, I get a 'label and input surfaces have different number of vertices' error which I assume is due to my gifti input file (which is the gifti version of the Glasser dlabel file, output by wb_command -cifti-separate) and my input surface (which I set as the individual subject's pial surface - please feel free to correct me here if I should be using a different input surface) having a different number of vertices (~32k and ~128k, respectively). Would you recommend that I resample the Glasser gifti label file into 128k using -label-resample before I run the  -label-to-volume-mapping command (or downsample the individual subject's input surface), and if so, which would be the appropriate sphere files to use here?

Thanks again!
Panos

Glasser, Matt

unread,
Oct 9, 2021, 4:45:16 PM10/9/21
to hcp-...@humanconnectome.org, tsc...@mst.edu

If you like you can resample the parcellation onto the native mesh.  There might be a minor improvement in accuracy.

Panos Fotiadis

unread,
Oct 9, 2021, 5:22:57 PM10/9/21
to HCP-Users, glas...@wustl.edu, tsc...@mst.edu

Sounds good, will do! Which file would you suggest that I use as my <current sphere> ( -> a sphere surface with the mesh that the label file is currently on) option for the -label-resample command? And, also, is the adap_bary_area method option of the command still recommended here?

Thanks again,
Panos

Glasser, Matt

unread,
Oct 10, 2021, 6:11:30 PM10/10/21
to hcp-...@humanconnectome.org, tsc...@mst.edu

You would use the 32k spheres for current and the MSMAll registered native mesh spheres for new.  adap_bary_area is correct.

Panos Fotiadis

unread,
Oct 12, 2021, 5:28:19 PM10/12/21
to HCP-Users, glas...@wustl.edu, tsc...@mst.edu
Perfect, thanks once again!

Cheers,
Panos

Reply all
Reply to author
Forward
0 new messages