Yes, it is a similar problem to working in "image space", basically the FSL
tools work in a space resulting from taking the voxel indices and
multiplying them by the voxel dimensions, so a particular corner voxel
always has coordinates of (0, 0, 0)mm. Because of the transition from
origin-less ANALYZE volume format to a completely defined coordinate space
in NIfTI, different software has ended up with different quirks as to
whether it properly handles NIfTI coordinates. I think ANTs does use the
full NIfTI coordinate specification, though internally they flip it to the
ITK convention (which only matters if you need to interpret ANTs transforms
yourself, as the identity transform still does the right thing).
Unfortunately, FSL continued to use the ANALYZE coordinate specification
when doing registration and resampling of NIfTI volumes.
We started connectome workbench much more recently than FSL started, and
although workbench inherited some code from caret5 (which also used the
full NIfTI coordinate specification, but also supported ANALYZE and other
formats, leading to some issues), its volume code was rewritten from
scratch, and only supports NIfTI volumes, so we were able to avoid legacy
issues and use the full NIfTI coordinates from day 1. FSL didn't have this
luxury, as they started before NIfTI even existed.
As to your situation, when using FSL tools to resample to a different voxel
grid, you have to either carefully generate your volume spaces in advance
to have their corners aligned in NIfTI space the same way that FSL assumes
they align in its ANALYZE-like space, or you have to deal with FSL adding
translations to your voxel coordinates every time you resample between
these voxel spaces (by using affine files containing the opposite
translations).
The different MNI template resolutions we use in the pipelines have these
special corner voxels aligned already in their NIfTI coordinates, so for
the pipelines we can tolerate the quirks of FSL resampling tools (and their
more advanced tools don't seem to have equivalents in other software
packages, and we work with them on improvements), but I can't recommend FSL
resampling tools to the unwary as the first choice for a new task in an
unknown volume space.
Tim
On Wed, May 29, 2019 at 1:16 PM Jaime Caballero <jcabai...@gmail.com> wrote:
> Thank you, Timothy
>
> I did the resampling with ANTs' WarpImageMultiTransform, as I thought the
> software used for this didn't matter. I have now done the same with
> wb_command as you suggested and the resulting ROIs are exactly the same.
>
> I could not make FSL's applywarp work, so I cannot tell by now what would
> come out from that. Anyway I'm not sure I understand why would it fail. Is
> it something related to rounding coordinates or working directly in image
> space?
>
> Regards,
> Jaime
>
> El mar., 28 may. 2019 a las 20:46, Timothy Coalson (<tsc...@mst.edu>)
> escribió:
>
>> If you need to preserve the mm coordinates of the ROI, I would not trust
>> FSL's resampling with an identity transform to get it right, because that
>> will produce a different shift depending on what coordinates a particular
>> corner voxel is at in each image (as I understand it, FSL's conventions
>> come from originally handling ANALYZE format images). Instead, wb_command
>> -volume-affine-resample, given an identity matrix (and NOT specifying
>> -flirt) will do the resampling via nifti mm coordinates, with no guesswork.
>>
>> Tim
>>
>>
>> On Tue, May 28, 2019 at 8:08 AM Jaime Caballero <jcabai...@gmail.com>
>> wrote:
>>
>>> Thank you so much!
>>>
>>> It worked fine, and the resulting ROIs are were they are suposed to be.
>>>
>>> Regards,
>>> Jaime
>>>
>>> El lun., 27 may. 2019 22:20, Glasser, Matthew <glass...@wustl.edu>
>>> escribió:
>>>
>>>> That file does exist in the structural package then.
>>>>
>>>>
>>>>
>>>> Matt.
>>>>
>>>>
>>>>
>>>> *From: *Jaime Caballero <jcabai...@gmail.com>
>>>> *Date: *Monday, May 27, 2019 at 3:20 PM
>>>> *To: *"Glasser, Matthew" <glass...@wustl.edu>
>>>> *Cc: *"hcp-users@humanconnectome.org" <hcp-users@humanconnectome.org>
>>>> *Subject: *Re: [HCP-Users] Convert nifti-ROIs to cifti format
>>>> (subcortical)
>>>>
>>>>
>>>>
>>>> Sorry for the confusion.
>>>>
>>>>
>>>>
>>>> Locally acquired data wasn't processed using the HCP pipelines, it is
>>>> in a different resolution and it was processed in volume space, no problem
>>>> there.
>>>>
>>>>
>>>>
>>>> I want to use Choi's parcellation with HCP data. All the process I
>>>> described is mi workaround to adapt Choi files to HCP files.
>>>>
>>>>
>>>>
>>>> Jaime
>>>>
>>>>
>>>>
>>>> El lun., 27 may. 2019 22:12, Glasser, Matthew <glass...@wustl.edu>
>>>> escribió:
>>>>
>>>> I thought you said you were using a locally collected sample you ran
>>>> the HCP Pipelines on?
>>>>
>>>> Matt.
>>>>
>>>>
>>>>
>>>> *From: *Jaime Caballero <jcabai...@gmail.com>
>>>> *Date: *Monday, May 27, 2019 at 3:10 PM
>>>> *To: *"Glasser, Matthew" <glass...@wustl.edu>
>>>> *Subject: *Re: [HCP-Users] Convert nifti-ROIs to cifti format
>>>> (subcortical)
>>>>
>>>>
>>>>
>>>> Ok, thank you! I will try that.
>>>>
>>>>
>>>>
>>>> I asume the reference image
>>>> ${StudyFolder}/${Subject}/MNINonLinear/ROIs/Atlas_ROIs.2.nii.gz is to be
>>>> downloaded with the structural package?
>>>>
>>>>
>>>>
>>>> Regards,
>>>>
>>>> Jaime
>>>>
>>>>
>>>>
>>>> El lun., 27 may. 2019 a las 21:00, Glasser, Matthew (<
>>>> glass...@wustl.edu>) escribió:
>>>>
>>>> To make the .dscalar.nii file, you seem to be on the right track. If
>>>> the Choi ROIs are properly in MNI space, hopefully you could simply use
>>>> applywarp --interp=nn -i <ROIs.nii.gz> -r
>>>> ${StudyFolder}/${Subject}/MNINonLinear/ROIs/Atlas_ROIs.2.nii.gz -o
>>>> <ROIs.2.nii> and then the wb_command -cifti-create-dense-from-template you
>>>> mention.
>>>>
>>>>
>>>>
>>>> Matt.
>>>>
>>>>
>>>>
>>>> *From: *Jaime Caballero <jcabai...@gmail.com>
>>>> *Date: *Monday, May 27, 2019 at 12:56 PM
>>>> *To: *"Glasser, Matthew" <glass...@wustl.edu>
>>>> *Cc: *"hcp-users@humanconnectome.org" <hcp-users@humanconnectome.org>
>>>> *Subject: *Re: [HCP-Users] Convert nifti-ROIs to cifti format
>>>> (subcortical)
>>>>
>>>>
>>>>
>>>> The objective is to extract functional connectivity between cortical
>>>> and striatal ROIs, and ALFF and ReHo from both cortical and striatal ROIs.
>>>>
>>>>
>>>>
>>>> Up to this point I have imported each subject's dtseries.nii file into
>>>> MATLAB, and also the previously defined ROIs in an HCP-compatible format.
>>>> For the dtseries I have a 96854x1200 matrix, and for the ROI a 96854x1
>>>> matrix containing a mask, which I use to extract the time series I'm
>>>> interested in for further processing.
>>>>
>>>>
>>>>
>>>> Jaime
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> El lun., 27 may. 2019 a las 19:14, Glasser, Matthew (<
>>>> glass...@wustl.edu>) escribió:
>>>>
>>>> What do you plan to do with the file?
>>>>
>>>>
>>>>
>>>> Matt.
>>>>
>>>>
>>>>
>>>> *From: *Jaime Caballero <jcabai...@gmail.com>
>>>> *Date: *Monday, May 27, 2019 at 12:08 PM
>>>> *To: *"Glasser, Matthew" <glass...@wustl.edu>
>>>> *Cc: *"hcp-users@humanconnectome.org" <hcp-users@humanconnectome.org>
>>>> *Subject: *Re: [HCP-Users] Convert nifti-ROIs to cifti format
>>>> (subcortical)
>>>>
>>>>
>>>>
>>>> A .dscalar.nii output, I think. Basically I want an equivalent of the
>>>> nifti ROI, but in cifti: for each voxel/vertex, value 1 if inside the ROI,
>>>> 0 if outside. Would a dlabel file be better for this application?
>>>>
>>>>
>>>>
>>>> El lun., 27 may. 2019 a las 19:03, Glasser, Matthew (<
>>>> glass...@wustl.edu>) escribió:
>>>>
>>>> Are you wanting a .dlabel.nii output or a .dscalar.nii output?
>>>>
>>>>
>>>>
>>>> Matt.
>>>>
>>>>
>>>>
>>>> *From: *<hcp-users-boun...@humanconnectome.org> on behalf of Jaime
>>>> Caballero <jcabai...@gmail.com>
>>>> *Date: *Monday, May 27, 2019 at 10:37 AM
>>>> *To: *"hcp-users@humanconnectome.org" <hcp-users@humanconnectome.org>
>>>> *Subject: *[HCP-Users] Convert nifti-ROIs to cifti format (subcortical)
>>>>
>>>>
>>>>
>>>> Dear experts
>>>>
>>>>
>>>>
>>>> In my center we are studying cortico-striatal functional connectivity,
>>>> and cortical/striatal local measures (ALFF, ReHo) on a locally acquired
>>>> sample. For that we are using Choi's functional parcellation, distributed
>>>> as a volumetric NIFTI file, in MNI152 space. We want to validate our
>>>> measures with a subset of the S1200 release (resting state, 3T).
>>>> Specifically I have used the ICA-FIX cleaned and MSM-all registered files,
>>>> i.e.:
>>>>
>>>>
>>>> <subject_dir>/MNINonLinear/Results/rfMRI_REST1_LR/rfMRI_REST1_LR_Atlas_MSMAll_hp2000_clean.dtseries.nii
>>>>
>>>>
>>>>
>>>> The thing is I have doubts on the way to convert nifti ROIs to cifti
>>>> format in this case.
>>>>
>>>>
>>>>
>>>> My first approach:
>>>>
>>>>
>>>>
>>>> 1. Load a cifti template file and the NIFTI ROI file in MATLAB. The roi
>>>> file is a volume containing 0s and 1s.
>>>>
>>>> 2. Use the inverse of the NIFTI's transformation matrix to convert the
>>>> XYZ coordinates in MNI space from each grayordinate to XYZ coordinates in
>>>> the NIFTI'S volume space. (The matrix is corrected to account for MATLAB's
>>>> 1-based matrix indexing)
>>>>
>>>> 3. The values of the NIFTI volume for the obtained coordinates define
>>>> the ROI in the output cifti.
>>>>
>>>> 4. Keep only the grayordinates that are set to one by my method and
>>>> that are labeled as striatum (i.e. caudate, putamen or accumbens) in the
>>>> cifti files.
>>>>
>>>>
>>>>
>>>> This aproach is rather hand-made, and I wonder if I am missing
>>>> something important. Conceptually it looks correct to me, but the ROIs
>>>> appear slightly displaced to the right, which might affect the results.
>>>>
>>>> To load the CIFTIs I use fieltdrip's ft_cifti_read( ), which for this
>>>> purpose works well (it reads the grayordinate positions and labels), and to
>>>> load the NIFTIs I use Freesurfer's load_nifti( ).
>>>>
>>>> Is this method correct, or is there something important I'm not taking
>>>> into account?
>>>>
>>>>
>>>>
>>>> Now I'm trying to do the same with wb_command to compare, but I cannot
>>>> get it working. The procedure I use is:
>>>>
>>>>
>>>>
>>>> # Resample the ROI file to 2x2x2 resolution with ANTs:
>>>>
>>>> ResampleImage 3 "$NiftiFileIn" "$NiftiFileResampled" 2x2x2 0
>>>>
>>>> # Convert the obtained nifti to cifti:
>>>>
>>>> wb_command -cifti-create-dense-from-template
>>>> "$CiftiDscalarTemplateFile" "$CiftiOut" -volume-all "$NiftiFileResampled"
>>>>
>>>>
>>>>
>>>> Which outputs the error: -volume-all specifies a volume file that
>>>> doesn't match the volume space of the template cifti file
>>>>
>>>> Is there anything wrong with my way of doing this procedure?
>>>>
>>>>
>>>>
>>>> Thanks in advance,
>>>>
>>>>
>>>>
>>>> Best regards,
>>>>
>>>> Jaime Caballero-Insaurriaga
>>>>
>>>> _______________________________________________
>>>> HCP-Users mailing list
>>>> HCP-Users@humanconnectome.org
>>>> http://lists.humanconnectome.org/mailman/listinfo/hcp-users
>>>>
>>>>
>>>> ------------------------------
>>>>
>>>> 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.
>>>>
>>>>
>>>> ------------------------------
>>>>
>>>> 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.
>>>>
>>>>
>>>> ------------------------------
>>>>
>>>> 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.
>>>>
>>>>
>>>> ------------------------------
>>>>
>>>> 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.
>>>>
>>>>
>>>> ------------------------------
>>>>
>>>> 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.
>>>>
>>> _______________________________________________
>>> HCP-Users mailing list
>>> HCP-Users@humanconnectome.org
>>> http://lists.humanconnectome.org/mailman/listinfo/hcp-users
>>>
>>
_______________________________________________
HCP-Users mailing list
HCP-Users@humanconnectome.org
http://lists.humanconnectome.org/mailman/listinfo/hcp-users
