How to relate surface-based activation map to volume-based data

[Narges Moradi]

Hello,

 I am converting surface_based coordinates of a task fMRI data from HCP to the volume_based coordinates using "wb_command -label-to-volume-mapping". How can I relate the "cope1.dtseries.nii" values of the surface-based coordinates to the volume-based coordinates? Thanks.

Summary: How can I have the activation map in volume-based task fMRI data?

[Glasser, Matthew]

Why would you want to do that?

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/289bf006-bfac-4e3c-ad64-79af38829070n%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.

[Coalson, Timothy Scott (S&T-Student)]

We generally recommend staying with surface-based files for cortical data, as it enables better cross-subject alignment.  Mapping between surface and volume is a slightly lossy process in individuals (and is usually a very lossy process for group data), so you should only do so when necessary.  What are you planning to do with a volume-based version of this file?

The -metric-to-volume-mapping command does the equivalent of -label-to-volume-mapping for real-valued data, but you would need to use -cifti-separate first, as the surface to volume mapping commands only handle one hemisphere at a time.

Tim

---

From: Narges Moradi <nargesmo...@gmail.com>
Sent: Friday, June 11, 2021 7:58 PM

To: HCP-Users <hcp-...@humanconnectome.org>
Subject: [hcp-users] How to relate surface-based activation map to volume-based data

Hello,

 I am converting surface_based coordinates of a task fMRI data from HCP to the volume_based coordinates using "wb_command -label-to-volume-mapping". How can I relate the "cope1.dtseries.nii" values of the surface-based coordinates to the volume-based coordinates? Thanks.

Summary: How can I have the activation map in volume-based task fMRI data?

--

[Narges Moradi]

I have a code for volume-based resting-state fMRI, and now I want to apply it to the task data. It would be easier and faster to have the data in the same format as the rfMRI data rather than changing the code for surface-based data. 

[Narges Moradi]

Thanks for the quick reply and explanation.
So you mean I can get the volume version of the "cope1.dtseries.nii"  using -metric-to-volume-mapping command?

I want the volume version because I have a code for volume-based resting-state fMRI, and now I want to apply it to the task data. It would be easier and faster to have the data in the same format as the rfMRI data rather than changing the code for surface-based data. 

[Narges Moradi]

Thanks for the quick reply and explanation.
So you mean I can get the volume version of the "cope1.dtseries.nii"  using -metric-to-volume-mapping command?

I want the volume version because I have a code for volume-based resting-state fMRI, and now I want to apply it to the task data. It would be easier and faster to have the data in the same format as the rfMRI data rather than changing the code for surface-based data. 

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/YNrBcg8XEDw/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/SN6PR01MB4991051378660632E6B80D1281339%40SN6PR01MB4991.prod.exchangelabs.com.

[Glasser, Matthew]

I would use the CIFTI formatted HCP resting state data, e.g. ${StudyFolder}/${Subject}/MNINonLinear/Results/${fMRIName}/${fMRIName}_Atlas_MSMAll_hp2000_clean.dtseries.nii.  There isn’t much point in using HCP data if you aren’t going to use HCP-Style analyses—you basically ruin the advantages of the data.  On the other hand, even traditional lower resolution data also benefits from HCP-Style analyses (see Coalson et al., 2018 PNAS). 

[Coalson, Timothy Scott (S&T-Student)]

Does the code you have do any smoothing or other spatial operations that need to know the neighbors of each data point?  If not, there is a workaround that would let you run the cifti data through your code without any damage due to converting to volume and back.  Even if it does have steps like that, you may be able to swap in our existing wb_command implementations of smoothing, gradient, and possibly other things without too much trouble.  Note that in particular, volume-based smoothing will cause cortical activations to "bleed" across sulci, which is one reason to avoid volume-based processing steps even for a single subject.

With cifti data, you don't need a separate ROI to exclude things that aren't gray matter, so you might find a cifti-based implementation to have fewer complications than a volume-based script.

Tim

---

From: Narges Moradi <nargesmo...@gmail.com>
Sent: Friday, June 11, 2021 11:03 PM
To: hcp-...@humanconnectome.org <hcp-...@humanconnectome.org>
Subject: Re: [hcp-users] How to relate surface-based activation map to volume-based data

 

To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/CAJgRpp5Qe7A71EFUztTo%2BLjrECkzmSrX2-kh4CsuG-91jMxFWA%40mail.gmail.com.

[Narges Moradi]

The code contains a 3D spatial decomposition of the fMRI image and knowing the neighbours of each data point is needed. I would like to use cifti data but the question is how to use the wb-command to obtain grey matter's coordinates in a 91*109*91 matrix format like in nifti data? Any help would be appreciated. Thanks.

[Coalson, Timothy Scott (S&T-Student)]

On the off chance that it is a "spatial ICA" decomposition, that doesn't actually need 3D relationships, and is something we already do with cifti data.

We would still recommend against using volume-based methods on cortical data, particularly for analyzing high resolution data.  As a secondary option, doing all volume-based steps in each single subject's volume data, without any steps that cause smoothing, and then mapping to the subject's surface before combining across subjects would eliminate the misalignment issues we are worried about (but requires the volume-based steps to be tolerant of noise).

I think our task pipelines can produce volume-based outputs for single subjects.  If the steps you need to do in the volume are done on single subjects, this may be the better option.  However, wb_command -cifti-separate and -metric-to-volume-mapping can be used to make a volume file from cifti data, though it will have some artifacts of having been surface-based data (background will be flat zeros with no randomness, edge of cortex will have a distinct dropoff, etc).

If the volume-based steps have to be run on group data, then you have a problem, as group-based volume data with current volume registrations still has trouble even lining up gray matter with gray matter, let alone aligning functional areas, and the resulting confusion of tissue identity makes it problematic to interpret (see our paper that Matt referenced).  There could be a workaround for this, by leveraging the good surface-based alignment in cifti format to make a "group average" volume based on surface alignment rather than volume registration (which could not be directly compared to actual volume-based results, and mapping it back to the surface would need yet another workaround), but really the proper solution is to adapt the methods to use the neighborhoods from the surface instead of from voxels.

Tim

---

From: Narges Moradi <nargesmo...@gmail.com>
Sent: Tuesday, June 15, 2021 1:39 AM
To: HCP-Users <hcp-...@humanconnectome.org>
Cc: Coalson, Timothy Scott (S&T-Student) <tsc...@mst.edu>

[Narges Moradi]

Thanks for the response, Tim. The decomposition in the volume is done on single subjects. 

Does the output volume file made by the "wb_command -cifti-separate" and "-metric-to-volume-mapping" include the subcortical voxels which are volume-based or I need to add them to the final volume_based data separately? 

Also, Is there any other command that can be used instead of "wb_command -merged-volume" to put all volume data (left and right hemisphere volumes) in one single volume? using this command I got the output volume of size 91*109*91*2 instead of 91*109*91*1. Thanks again.

[Coalson, Timothy Scott (S&T-Student)]

No, -metric-to-volume-mapping operates only on surface data, the subcortical data would need to be extracted and added separately.  As long as you are doing this part of the analysis in single subjects (and aren't doing smoothing), it may be easier to use volume-based task activations than to reverse engineer them from the cifti data (and then map the cortical results onto the subject's surfaces before comparing across subjects).

Combining the mapped to volume data would need to be done with -volume-math, and might be tricky depending on how you want to handle overlaps between hemispheres (and subcortical structures).

Tim

---

From: Narges Moradi <nargesmo...@gmail.com>
Sent: Tuesday, June 22, 2021 12:39 AM

[Narges Moradi]

Thanks, Tim. Would you please explain what do you mean by "volume-based task activations", Is it another hcp dataset? and is it preprocessed data?

[Glasser, Matthew]

I don’t see anywhere in this thread where you have told us specifically what you are trying to do (i.e. why you are asking this question).  Those kind of details (off list if needed) would be helpful to give you the best possible advice.  That said, it is unlikely we are going to advise you to do traditional volume-based analyses.  If you are stuck in that world, it will be hard to make optimal use of HCP data (and it won’t really offer you benefits over legacy data in many respects).  Indeed even low resolution legacy data does better with HCP-Style approaches to preprocessing and data analysis than with traditional approaches.  They are causing major ongoing problems with reproducibility and neuroanatomical localization in the brain imaging field that aren’t really solvable using traditional approaches. 

 

Matt.

 

From: Narges Moradi <nargesmo...@gmail.com>

Reply-To: "hcp-...@humanconnectome.org" <hcp-...@humanconnectome.org>
Date: Tuesday, June 22, 2021 at 1:26 PM
To: HCP-Users <hcp-...@humanconnectome.org>

To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/a604b3ef-219e-4ca7-b0c0-953449b9d747n%40humanconnectome.org.

[Coalson, Timothy Scott (S&T-Student)]

I think you would need to run level1 (single run) and level2 (multi-run, single subject) task analysis with smoothing set to 0 and volume outputs enabled, as I don't think we have done this (hopefully your code can handle unsmoothed single subject data).  Note that level3 (cross-subject) should not be done with cortical data in volume form.

Can you give some details on what type of decomposition you are doing (off-list if you prefer)?  It would be good to know how difficult it would actually be to adapt it to cifti data, since mapping between volume and surface is nontrivial and somewhat lossy (it is effectively a resampling of sorts, and thus causes some slight blurring each time), while surface-based methods allow you to do smoothing that doesn't cross between opposing sulcal banks or dilute into white matter or csf.

Tim

---

From: Narges Moradi <nargesmo...@gmail.com>
Sent: Tuesday, June 22, 2021 1:26 PM
To: HCP-Users <hcp-...@humanconnectome.org>
Cc: Coalson, Timothy Scott (S&T-Student) <tsc...@mst.edu>; Narges Moradi <nargesmo...@gmail.com>

[Glasser, Matthew]

Tim mentioned that this was an individual subject analysis.  That can be okay in the volume as long as you don’t smooth the data, but it would still be helpful to know exactly what you are trying to do.

Matt.

To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/ECC99468-ABBE-43D0-9186-8FE5C84B0DEF%40wustl.edu.