Label files for diffusion and fMRI

397 views
Skip to first unread message

Shael Brown

unread,
Apr 22, 2021, 8:47:18 AM4/22/21
to HCP-Users
Hi folks,

I'm trying to perform an ROI study linking diffusion data and fMRI data within subjects, and then comparing the results between subjects in the 1200 subjects release 7T data. Here are my questions:

(1) The diffusion data is easy to find, however I see two possible (rs) fMRI files: for example in the same directory I see rfMRI_REST1_7T_PA_Atlas_1.6mm.dtseries.nii and rfMRI_REST1_7T_PA_Atlas_1.6mm_MSMAII.dtseries.nii. What is the difference between the two? 

(2) I'm looking for label maps with a good number of areas (at least about 100) for both the diffusion and fMRI data, such that area label "1" in the diffusion label map is the same area label "1" in the fMRI label map, and there's somewhere that I can look up the name of area label "1" (for ex "V1" etc). Where could I find such files?

(3) Will the rs-fMRI label map also apply to the task-based fMRI data or is there a new file?

Thanks in advance!!

Best,
Shael

Glasser, Matthew

unread,
Apr 22, 2021, 2:32:14 PM4/22/21
to hcp-...@humanconnectome.org
  1. The former is aligned across subjects with MSMSulc (folding) and the latter is aligned across subjects with MSMAll (areal features).  I believe it may not yet have been demonstrated whether diffusion tractography results (if that is what you are referring to) follows folding or cortical areas more.  Cortical diffusion properties (e.g. NODDI) follow cortical areas.  Unless you have a specific hypothesis requiring 1.6mm data, I would stick with the 2mm data as it is half the size and already quite high resolution (and more compatible with a wider set of files). 
  2. It sounds like you are in need of a cortical parcellation.  I recommend the HCP’s multi-modal parcellation: https://balsa.wustl.edu/file/show/3VLx
  3. I am of the opinion that there is a single set of cortical areas in the brain that are consistent across multiple modalities.  I do not agree with the view that every modality has its own parcellation.  It is the case that no modality will reveal all cortical areal boundaries, and therefore it is the complementary and converging evidence across multiple modalities that gives the best parcellation.


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/4e4b2259-9573-43cb-9002-3d7bf3b225dbn%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.

Shael Brown

unread,
Apr 23, 2021, 7:55:05 AM4/23/21
to HCP-Users, glas...@wustl.edu
Hi Matt,

Thanks for the quick reply. A few follow-up questions:

1. Indeed I am referring to tractography results for diffusion. I only see 1.60mm resolution in the 7T data, so are you suggesting that 3T data would be preferable? In the 3T subject files I downloaded it looks like there are two folders containing resting-state data, rfMRI_REST1_RL and rfMRI_REST1_LR - what is the difference?

2. Yup a parcellation would be great! I checked out what you sent and it looks like it's a surface parcellation? I'm wanting to include subcortical vertices too (grayordinates?) so a volume parcellation would be ideal. Since everything is registered into MNI space for all subjects (for both diffusion and fMRI data I believe?) would I need individual parcellations for each subject (possibly one for each of diffusion and fMRI) or one for everyone including diffusion and fMRI data?

3. I'd agree with that!

Thanks,
Shael

Glasser, Matthew

unread,
Apr 23, 2021, 8:27:04 AM4/23/21
to Shael Brown, HCP-Users
  1. There are both 2mm and 1.6mm 7T CIFTI fMRI data.  The 2mm data doesn’t say 2mm though.
  2. You can map that parcellation to the individual’s volume space using their own surfaces.  You would not want to use a volume-based group average cortical parcellation due to the issues described in Glasser et al., 2016 Nature Neuroscience and Coalson et al., PNAS).

Shael Brown

unread,
Apr 23, 2021, 8:57:21 AM4/23/21
to Glasser, Matthew, HCP-Users
1. Perfect, thanks!
2. That sounds like exactly what I'm looking for. Do you know of a link to a workflow to accomplish this?

Best,
Shael

Coalson, Timothy Scott (S&T-Student)

unread,
Apr 23, 2021, 4:40:03 PM4/23/21
to Glasser, Matthew, hcp-...@humanconnectome.org
We would suggest that cortical-focused tractography use surfaces for counting hits, as voxel-based masks create jagged "surfaces" that partially poke into white matter, and properly utilizing the values from the resulting thin layer of counting voxels could be tricky.  There are other biases we have found in tractography (particularly in estimating cortical connectivity, the "gyral bias") that I don't think have been resolved yet, so you may need to plan around having limitations in accuracy.  With hits counted on the surface, you would want to use a surface-based parcellation anyway rather than converting to voxels.  If you don't choose this route, mapping surface label data to voxels can be done with wb_command -label-to-volume-mapping.

You can also combine a subcortical parcellation with the surface-based MMP in cifti format without converting any surface data to voxels, and that could be used directly on our fMRI cifti files.  I would suggest wb_command -cifti-create-dense-from-template for this, use any standard 91282 grayordinate file as the template.  You will need to ensure that the subcortical parcellation uses the same exact voxel grid as our 2mm data for the command to accept it.

Tim

From: Shael Brown <shael...@gmail.com>
Sent: Friday, April 23, 2021 7:57 AM
To: Glasser, Matthew <glas...@wustl.edu>
Cc: HCP-Users <hcp-...@humanconnectome.org>

Shael Brown

unread,
Apr 26, 2021, 9:38:30 AM4/26/21
to HCP-Users, Glasser, Matthew
Thanks for the response, Tim. I'm still lost so I'll try to be even more specific. I downloaded one subject's anatomical, diffusion and rs-fMRI data and these are my updated questions:

- In the MNINonLinear/ directory I have a folder called ROIs in which there's the file Atlas_wmparc.1.60.nii which has about 180 labels. Is this the Desikan-Killany atlas? And where would I find the region names for each label number? 
- Finally, is there a way to determine for each diffusion voxel (grayordinate?) which atlas region from the Atlas_wmparc.1.60.nii file it belongs to if any? The diffusion data is sampled at 1.05mm voxels, so perhaps there would have to be some resampling involved? I don't see any label files in the anatomical folders.

Thanks so much,
Shael

To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/SN6PR01MB4991C3090067A77FCC31729481459%40SN6PR01MB4991.prod.exchangelabs.com.

Glasser, Matthew

unread,
Apr 26, 2021, 10:03:16 AM4/26/21
to Shael Brown, HCP-Users

As Tim says, we wouldn’t use these files for this purpose.

Coalson, Timothy Scott (S&T-Student)

unread,
Apr 26, 2021, 5:34:33 PM4/26/21
to HCP-Users, Glasser, Matthew
That is still a volume file, not a surface file (.surf.gii), so we would not recommend using anything derived from it for the counting of cortical tractography hits.  We are talking about something like the --mesh option in probtrackx.  If you were only looking within deep white matter (where you don't need to count endpoints), wmparc might be useful.

That wmparc volume has been imported to workbench label format, so -volume-label-export-table will work on it.  However, it was likely downsampled from the original freesurfer output, which would amplify the issues with it.

Tim

From: Shael Brown <shael...@gmail.com>
Sent: Monday, April 26, 2021 8:38 AM

Coalson, Timothy Scott (S&T-Student)

unread,
Apr 26, 2021, 5:37:29 PM4/26/21
to HCP-Users, Glasser, Matthew
Sorry, missed an issue: since it is inside MNINonLinear, that wmparc file is probably no longer the size/shape of the subject's head/brain, either.

Tim

From: Coalson, Timothy Scott (S&T-Student) <tsc...@mst.edu>
Sent: Monday, April 26, 2021 4:34 PM

Shael Brown

unread,
Apr 30, 2021, 10:07:22 AM4/30/21
to HCP-Users, Glasser, Matthew
Hi everyone,

Thanks for the responses. I'm wanting to use the workbench and the multimodal HCP parcellation which I've downloaded to label individual subject diffusion and fMRI volume datasets. Here are the specific questions I have:

(1) Which is the correct fMRI file I should be using for each subject? Inside each folder like rfMRI_REST1_LR in MNINonLinear/Results/ there are many files, but would the minimally preprocessed file which I can use to warp to the multimodal parcellation be rfMRI_REST1_LR.nii or a different file in that folder?

(2) I have found the documentation for the workbench command -cifti-create-dense-from-template at https://www.humanconnectome.org/software/workbench-command/-cifti-create-dense-from-template but I would immensely appreciate a quick example of how to run the command from windows powershell. I would imagine it would be something like this:

- for fMRI:   path_to_wb/wb.exe -series -template-cifti path_to_multimodal_parcellation/parc.nii -cifti-out output_path/fMRI_parc.nii  -label path_to_multimodal_parcellation/parc.nii -volume-in path_to_fMRI_file/fMRI.nii

- for dMRI:  path_to_wb/wb.exe -series -template-cifti path_to_multimodal_parcellation/parc.nii -cifti-out output_path/diffusion_parc.nii  -label path_to_multimodal_parcellation/parc.nii -volume-in path_to_diffusion_file/diffusion_data.nii

But I really could use some assistance here. Again I'm trying to attach a label from the multimodal parcellation to each voxel (cortical and subcortical) in both the fMRI file and the diffusion file.

(3) Where could I find familial relations (siblings or twins) between the subject IDs?

Thank you all so much for your help.

Best,
Shael

To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/SN6PR01MB4991698212D8C0FD38387ABF81429%40SN6PR01MB4991.prod.exchangelabs.com.

Glasser, Matthew

unread,
Apr 30, 2021, 10:11:02 AM4/30/21
to hcp-...@humanconnectome.org

(1-2) Are you sure you shouldn’t using the CIFTI fMRI data and mapping the diffusion data to the surface or using surface-based tractography?  It is rare that the optimal approach is volume-based, but in select situations it could make sense to map the parcellation to individual volumes using individual surfaces.

(3) Your PI would need to apply for restricted data access.  Instructions are available on the HCP website.

Elam, Jennifer

unread,
Apr 30, 2021, 10:12:26 AM4/30/21
to HCP-Users, Glasser, Matthew
Hi Shael,
For #3, the relatedness information--- family IDs etc. are Restricted Access data. Have you applied for and received HCP-YA Restricted access yet?

Jenn
From: Shael Brown <shael...@gmail.com>
Sent: Friday, April 30, 2021 9:07 AM

To: HCP-Users <hcp-...@humanconnectome.org>
Cc: Glasser, Matthew <glas...@wustl.edu>
Subject: Re: [hcp-users] Label files for diffusion and fMRI
 

* External Email - Caution *

Damion V Demeter

unread,
Apr 30, 2021, 10:27:21 AM4/30/21
to hcp-...@humanconnectome.org
Hey Shael, 

Unless things have changed recently, the demographic information you can download from the NDAR site will tell you which participants are twins. (I believe it's the acspsw03 file off the top of my head, but I could be remembering incorrectly.)  However, it doesn't tell you if they are monozygotic or dizygotic. You need to download the genetic info and use plink to calculate this. I found a website that walked through the process, which I followed about a year ago, but I am not able to find it right now. Hopefully that helps get you toward the right path for question #3. 

Damion

-----
Damion Demeter, MA
Doctoral Candidate
UT Austin | Cognitive Neuroscience


Harms, Michael

unread,
Apr 30, 2021, 10:37:10 AM4/30/21
to hcp-...@humanconnectome.org

 

Two comments to avoid any confusion:

The HCP-YA data (which is what we are referring to here) is not currently available on the NDA (which is where the HCP-Lifespan data is hosted).

Jenn can correct me if I’m wrong, but I think that the Restricted Access HCP-YA data includes a field with genetically confirmed zygosity.  No need to download the genetic data and compute that your own.

 

Cheers,

-MH

-- 

Michael Harms, Ph.D.

-----------------------------------------------------------

Associate Professor of Psychiatry

Washington University School of Medicine

Department of Psychiatry, Box 8134

660 South Euclid Ave.                        Tel: 314-747-6173

St. Louis, MO  63110                          Email: mha...@wustl.edu

 

From: Damion V Demeter <dem...@utexas.edu>
Reply-To: "hcp-...@humanconnectome.org" <hcp-...@humanconnectome.org>
Date: Friday, April 30, 2021 at 9:27 AM
To: "hcp-...@humanconnectome.org" <hcp-...@humanconnectome.org>
Subject: Re: [hcp-users] Label files for diffusion and fMRI

 

* External Email - Caution *

Hey Shael, 

Elam, Jennifer

unread,
Apr 30, 2021, 10:40:12 AM4/30/21
to hcp-...@humanconnectome.org
Hi Damion,
I think Shael is talking about HCP Young Adult (HCP-YA) subjects, and that data is NOT in the NDA.  However, Damion, you may be referring to the HCP-YA genetic data in dbGaP. If that is the case then I think you might be correct about the file to look through. 

On the other hand, Shael may not need all the genetic data and going through the hoops to get access to dbGaP would be overkill to get the family info that is available as Restricted data in ConnectomeDB, when one applies for Restricted Access using the form here: https://www.humanconnectome.org/storage/app/media/data_use_terms/DataUseTerms_HCP_RestrictedAccess_30Nov2017.pdf

Also, to be clear, the Lifespan subjects that ARE released in NDA were NOT recruited as twins/siblings-- there are some parents/kids/grandkids and a few siblings in the HCP-Aging and HCP-Development cohorts, but not really very many family connections overall.

Best,
Jenn

From: Damion V Demeter <dem...@utexas.edu>
Sent: Friday, April 30, 2021 9:27 AM
To: hcp-...@humanconnectome.org <hcp-...@humanconnectome.org>

Subject: Re: [hcp-users] Label files for diffusion and fMRI
 

* External Email - Caution *
Hey Shael, 

Damion V Demeter

unread,
Apr 30, 2021, 12:26:34 PM4/30/21
to hcp-...@humanconnectome.org
Oh, yeah, it was my mistake. I thought this was referring to the ABCD data on the NDA site. Sorry about that.

Damion

-----
Damion Demeter, MA
Doctoral Candidate
UT Austin | Cognitive Neuroscience

To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/CH2PR02MB64914222C81E1C2613C2E03ACC5E9%40CH2PR02MB6491.namprd02.prod.outlook.com.

Shael Brown

unread,
Apr 30, 2021, 3:47:10 PM4/30/21
to HCP-Users
Thanks for the tips folks! I'll fill out the form and apply for the restricted data.

To answer your question Matt for the moment I would still like to perform volume analyses, so my questions remain for points (1) and (2). Let me know if anyone has any suggestions! I'm recopying them here:

(1) Which is the correct fMRI file I should be using for each subject? Inside each folder like rfMRI_REST1_LR in MNINonLinear/Results/ there are many files, but would the minimally preprocessed file which I can use to warp to the multimodal parcellation be rfMRI_REST1_LR.nii or a different file in that folder?

(2) I have found the documentation for the workbench command -cifti-create-dense-from-template at https://www.humanconnectome.org/software/workbench-command/-cifti-create-dense-from-template but I would immensely appreciate a quick example of how to run the command from windows powershell. I would imagine it would be something like this:

- for fMRI:   path_to_wb/wb.exe -series -template-cifti path_to_multimodal_parcellation/parc.nii -cifti-out output_path/fMRI_parc.nii  -label path_to_multimodal_parcellation/parc.nii -volume-in path_to_fMRI_file/fMRI.nii

- for dMRI:  path_to_wb/wb.exe -series -template-cifti path_to_multimodal_parcellation/parc.nii -cifti-out output_path/diffusion_parc.nii  -label path_to_multimodal_parcellation/parc.nii -volume-in path_to_diffusion_file/diffusion_data.nii

But I really could use some assistance here. Again I'm trying to attach a label from the multimodal parcellation to each voxel (cortical and subcortical) in both the fMRI file and the diffusion file.
You received this message because you are subscribed to a topic in the Google Groups "HCP-Users" group.
To unsubscribe from this topic, visit https://groups.google.com/a/humanconnectome.org/d/topic/hcp-users/7co1s05E_9Y/unsubscribe.
To unsubscribe from this group and all its topics, 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/CAEb2v%2Brbn%3Dby4C%3DL4vmuhvYBQS%3DWMyX7iB0G90BqBrmg%3DsawPA%40mail.gmail.com.

Coalson, Timothy Scott (S&T-Student)

unread,
Apr 30, 2021, 5:06:12 PM4/30/21
to HCP-Users
Take a look at the output of wb_command -arguments-help.  Here is how you would combine the HCP MMP v1.0 with a subcortical parcellation (that has already been through -volume-label-import - note that you should try to avoid colliding label key values between the files you intend to combine) as a cifti file:

<path/to/bin>/wb_command -cifti-create-dense-from-template <path/to/pipelines>/global/templates/91282_Greyordinates/91282_Greyordinates.dscalar.nii combined.dlabel.nii -cifti <another/path>/MMP1.0.dlabel.nii -volume-all <yet/another/path>/subcortical.label.nii.gz

Powershell appears to not need backslashes anymore, and I would expect it doesn't need ".exe" either, so it can probably use basically the same line as on linux/mac.

Once you have this, you can parcellate any standard dense cifti file:

<path/to/bin>/wb_command -cifti-parcellate <some/path>/rfMRI_concat_MSMAll.dtseries.nii <anoter/path>/combined.dlabel.nii COLUMN rfMRI_concat_MMP.ptseries.nii

Note, these are not volume files.  If you want to map cortical data back to voxels (for instance, if you plan to stop at single-subject analysis), you need to use wb_command -cifti-separate to convert to gifti files for each hemisphere, plus the subcortical voxel data, then -metric-to-volume-mapping and/or -label-to-volume-mapping to turn the gifti files into separate per-hemisphere volume files, and then you can combine them into a single volume file, probably involving -volume-math.

FYI, most of our scripts are in bash, and we generally recommend linux for processing, partly because most computing clusters run linux.

Tim


From: Shael Brown <shael...@gmail.com>
Sent: Friday, April 30, 2021 2:46 PM
To: HCP-Users <hcp-...@humanconnectome.org>

Shael Brown

unread,
May 3, 2021, 7:59:49 AM5/3/21
to HCP-Users
Hi Tim,

Thanks for getting back to me. This seems like what I'm looking for, so thank you! Since this has multiple steps and I don't want to keep spamming you all with emails, is there any chance I could borrow a little bit of your time on zoom (sometime this week?) to make sure I'm doing everything properly? I know you're probably very busy but it would be a huge help to me. Let me know!

Thanks,
Shael

To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/BL0PR01MB49804E736D181D6ECEEBE427815E9%40BL0PR01MB4980.prod.exchangelabs.com.
Reply all
Reply to author
Forward
0 new messages