Ciftify output and correlation

[StephDocTUM]

Hi,

As you mentioned you use data which has been preprocessed with FMRIPREP. But in FMRIPREP no denoising is done. So is this step done somewhere in the ciftify commands?

Do I understand it right that this file ses-1_task-rest_Atlas_s0.dtseries.nii contains the time series from surface (in a 32k mesh?) and from all volume structure?

I would like to calculate the correlation matrices for only the surface of right hemisphere to rh, lh to lh, lh to rh, rh to lh and then further calculate things with these correlation values. (So actually I need to *.dconn.nii files but I could not find any in the output?)

I would first downsample the 32k mesh(ses-1_task-rest_Atlas_s0.dtseries.nii) to 4k with wb_command -cifti-resample Then split the data into lh cortex and rh cortex with wb_command -cifti-separate into cortex left and right And then use -cifti-pairwise-correlation to calculate my correlations. I would then further calculate my stuff and keep the indices of my correlation values. In the end I would have a 1D matrix (with one value for each vertex). How can I map these values back to surface then?

Best, Stephan

[edickie]

Hi Stephan,

1. ciftify does not do any cleaning in the pipeline because, like fMRIprep, we want to keep the decision of what nuisance signals to remove up to the end user. There is ciftify_clean_img as a helper for cleaning - that can read the confound tsvs from fmriprep.

2. To map calculated values back to cifti files - I would check out wb_command -cifti-convert -from-text as well as wb_command -create-dense-from-template. There is also now some useful tools in nibabel for reading and writing directly in python.

I hope that helps..

[StephDocTUM]

Thanks a lot! The output of fmriprep is RAS orientation, but I would need to change that into LAS orientation and give ciftify as input the LAS oriented images. Do you know if ciftify changes somehow the orientation in the algorithm? (register to RAS oriented templates etc.)?

Best, Steph

[edickie]

the python code in ciftify doesn't do anything to the LAS vs RAS orientation. But the dependencies (FSL and connectome-workbench) might?

[tiborauer]

I would first downsample the 32k mesh(ses-1_task-rest_Atlas_s0.dtseries.nii) to 4k with wb_command -cifti-resample

Hi @StephDocTUM, I understand it is an old issue, but I hope you can explain how you downsample CIfTI.