Nifti files to be used with HCP Workbench

[shilpitju]

Hi,

I have preprocessed my fMRI data using fMRIPrep with fs-no-reconall. My output functional and structural files are in nii format. However, later I realized I want to use 'The Cole-Anticevic Brain-wide Network Partition (CAB-NP)' (that is in Cifti format), for further ROI based analysis. This atlas can be accessed using HCP Workbench.

Can you please suggest me a way out to solve this issue?

Thanks,

Shilpi Modi

[effigies]

Add fsLR to your output spaces to get CIFTI outputs. Also, CIFTI files can only be made if the surfaces are reconstructed. --fs-no-reconall will prevent this.

[shilpitju]

Alright. Thank you.

[shilpitju]

Hi,

Is --cifti-output different from fsLR?

[mgxd]

hi @shilpitju - ---cifti-output is the flag necessary to produce your BOLD runs as CIFTI dtseries. Adding fsLR to your --output-spaces is not necessary.

However, I would strongly recommend waiting to produce CIFTI output until a new release (which will be 20.0.3) - the current implementation is producing incorrect data.

[shilpitju]

Thanks.

When do we expect the new release?

[mgxd]

We should have one out later today

[shilpitju]

That's fantastic. Thanks

[shilpitju]

Please have a look on my command below:

fmriprep-docker .\inputdir .\outputdir participant --participant-label sub-0001 --fs-lic
ense-file .\license.txt --use-syn-sdc --output-spaces MNI152NLin6Asym MNI152NLin2009cAsym fsLR --cifti-output --stop-on-first-crash

Will it serve my purpose of obtaining output in nifti as well as cifti formats?

[effigies]

Do you actually want all of these outputs? What spaces would you like to do analysis in?

[shilpitju]

For nii, I want my outputs in MNInonlinearv6 space. However, I think SDC only works when MNI152NLin2009cAsym is chosen as an output space (M I wrong here?). Plus, for Cifti fsLR and --cifti-output have to be used.

[effigies]

In 20.0.x, you should not have to adjust output spaces for particular processing options. You can drop the 2009cAsym; it will be used internally but won't be placed in your outputs. Similarly, according to @mgxd, you don't need fsLR, just --cifti-output. So:

fmriprep-docker inputdir outputdir participant \
    --participant-label 0001 --fs-license-file license.txt \
    --use-syn-sdc --output-spaces MNI152NLin6Asym \
    --cifti-output --stop-on-first-crash

[shilpitju]

Thanks a ton !!!

[shilpitju]

One more question plz, in order to run SDC, do we need to choose --output-spaces MNI152NLin6Asym (even if we don't want output in nifti)?

[effigies]

No.

[shilpitju]

Thanks again.

[shilpitju]

When is 20.0.x release expected?

[effigies]

Pending #2032. Hopefully today. Possibly tomorrow.

[shilpitju]

All the best.

[shilpitju]

fmriprep-docker inputdir outputdir participant
--participant-label 0001 --fs-license-file license.txt
--use-syn-sdc --cifti-output --stop-on-first-crash

In the above command since I have not given any option for '--output-spaces', will the output be only in Cifti or MNI as well?

[oesteban]

That command line will only generate CIFTI2 grayordinates files.

[effigies]

Do we not still default to 2009c if output spaces is empty?

[mgxd]

We do, 2009c outputs should be there if --output-spaces isn't defined

[oesteban]

Sorry, @shilpitju, as @effigies points out, I'm wrong. With such a command line, fMRIPrep should give you both types of outputs: one CIFTI2 grayordinates file and one NIfTI volume one the default MNI152NLin2009cAsym space.

If you don't want the volumetric outputs to be written, you can add the --output-spaces flag without any arguments, e.g.:

fmriprep-docker inputdir outputdir participant \
    --participant-label 0001 --fs-license-file license.txt \
    --use-syn-sdc --cifti-output --stop-on-first-crash \
    --output-spaces

[shilpitju]

Thanks

[shilpitju]

Hi,

I tried to use -cifti-parcellate command to parcellate my input cifti functional file using 'The Cole-Anticevic Brain-wide Network Partition'.

I used the fillowing command:

eval(['!wb_command -cifti-parcellate ' inputFile ' ' parcelCIFTIFile ' COLUMN ' parcelTSFilename ' -method MEAN'])

The output obtained is as:

While running:
wb_command -cifti-parcellate input.dtseries.nii CortexSubcortex_ColeAnticevic_NetPartition_wSubcorGSR_parcels_LR.dlabel.nii COLUMN Output_Atlas_CortSubcort.Parcels.LR.ptseries.nii -method MEAN

ERROR: input cifti files must have the same volume space

My input dtseries was obtained after processing using fmriprep 20.0.3

Regards,

Shilpi Modi

[shilpitju]

In continuation to above are volume-based subcortical voxels included in fmriprep preprocessing? Intuitively they should be, but just wanted to confirm.

Thanks

[mgxd]

hi @shilpitju - yes, volumes are present in the files. I took a deeper look into this, here's what I found:

• Inspecting the volumes of some CIFTI files generated with HCPPipelines, the dimensions match the files produced by fmriprep. However, HCP volumes are in LAS orientation, while fmriprep produces RAS. Producing an image in LAS orientation fixes the above error, though -cifti-parcellate should probably be expanded to handle RAS orientation as well.

• When generating the subcortical structures, fmriprep assumes the label file orientation matches the BOLD - currently this isn't the case, so the subcortical rois may be slightly off. There's a fix underway in BF: Ensure BOLD and label orientations are equal niworkflows#477 that will ensure these files match, and subsequently produce a volume in LAS for easier portability with workbench tools.

[shilpitju]

Thanks. I will look into nipreps/niworkflows#477 . However, can ciftify tools sort out these issues?

[oesteban]

paging @edickie

[edickie]

It sounds like the volumn output is in the cifti file, but it may not be the same volume space as teh fsLR outputs from HCPPipelines (or ciftify)? How are you resampling the volume output before creating the cifti files?

@shilpitju A fast workaround for is to use "wb_command -cifti-separate" to separate the Antecivic-Cole Atlas into separate Left-cortex (gifti), right-cortex (gifti) and subcortical atlas files. The subcortical atlas ROIs would be in MNI space (nifti mask)... so you could easily use them to extract timeseries from the MNI space nifti fmriprep outputs...

[shilpitju]

@edickie I have not yet used ciftify tools (have just pre-processed the data using fmriprep) . I was thinking that since freesurfer has already been run on my structural data, so now ciftify can be used to get everything into cifti.

However, as suggested by you if I separate the cole atlas into left-right-cortex and subcortex atlas files, will it sort the issue of LAS, RAS mismatch between fmriprep cifti output and HCP cifti?

[mgxd]

@shilpitju it won't, but you can reorient the volume and then recombine the files to generate the CIFTI.

Alternatively, you can rerun using the new release that will be out later today (20.0.4) which should allow you to parcellate without raising an error.

[shilpitju]

Alright. Thanks

[mgxd]

20.0.4 is now available - please reopen the issue persists

[shilpitju]

Hi,

I ran a dataset today morning using the latest version, unfortunately while trying to parcellate I m getting the same error.

Shall I share the dataset and the code I m using with you?

[mgxd]

@shilpitju is it possible you are reusing the same working directory? You may need to clear it.

[shilpitju]

No, I made a new one. I copied the newly generated cifti files to the workbench folder. That folder also contained the Cole-Atlas files used in the code.

[shilpitju]

Hi,

What are your suggestions to me now based on the reply in
CIFTI volume preferred orientation #39

Shall I use wb_command -cifti-resample or wait for an update in fmriprep?