Ciftify_clean_image

[shilpitju]

@edickie I downloaded ciftify using pip install ciftify to be run on windows subsystem for linux. I have my fmri data preprocessed using fmriprep and obtained as a cifti dtseries file. I want to perform nuissance regression of motion regressors and few other covariates using cifti_clean_img.

The command I gave was:

ciftify_clean_img --output-file=/mnt/e/SentCompData_Paper/Derivatives/fmriprep/sub-LTLE29/ses-1/func/sub-LTLE29_denoised_withmotion_acompcorr_ciftify_ses-1_task-SC_acq-TR2500_space-fsLR_den-91k_bold.dtseries.nii --confounds-tsv=/mnt/e/SentCompData_Paper/Derivatives/fmriprep/sub-LTLE29/ses-1/func/sub-LTLE29_ses-1_task-SC_acq-TR2500_desc-confounds_regressors.tsv --cf-cols=a_comp_cor_00 a_comp_cor_01 a_comp_cor_02 a_comp_cor_03 a_comp_cor_04 a_comp_cor_05 cosine00 cosine01 cosine02 trans_x trans_x_derivative1 trans_x_power2 trans_x_derivative1_power2 trans_y trans_y_derivative1 trans_y_derivative1_power2 trans_y_power2 trans_z trans_z_derivative1 trans_z_derivative1_power2 trans_z_power2 rot_x rot_x_derivative1 rot_x_derivative1_power2 rot_x_power2 rot_y rot_y_derivative1 rot_y_derivative1_power2 rot_y_power2 rot_z rot_z_derivative1 rot_z_power2 rot_z_derivative1_power2 --func_input=/mnt/e/SentCompData_Paper/Derivatives/fmriprep/sub-LTLE29/ses-1/func/sub-LTLE29_ses-1_task-SC_acq-TR2500_space-fsLR_den-91k_bold.dtseries.nii --debug --verbose

The command didn't get executed, however python did not give any error. Can you please help me running the command.

Thanks and regards,

Shilpi Modi

[shilpitju]

Hi,

I realized that the mistake on my part was not giving commas between the regressor columns to be selected. The command ran after doing so.

However,

1. the output denoised file is reduced in size. (86192 KB as a fmriprep output cifti file to 43406 KB after ciftify_image_clean)

2. Can I add few other covariates as regressors such as linear trend in the data, task effects etc?

Thanks

[edickie]

The reduced size is odd - I don't know it one of the step did this by changing the encoding? (i.e. from 32 to 16 bit?). Are the dimensions of the image still the same?

You could add additional things like task effects by adding them to the "confounds" tsv file..

There is also a "--detrend" option that can be added to the command line (or the config json) to allow for a linear detrend.

Hope that helps, Erin

[shilpitju]

Yes, the image dimensions are same.

I came to know from the HCP mailing list that conversion to nifti does not include the cifti XML at all, and if you don't use -smaller-file or -smaller-dims, there can be nontrivial wasted space (filled with zeros) compared to the size of just the data matrix. Further, -cifti-convert -to-nifti can write directly to a .nii.gz file to save some disk space. There could have been a significant amount of metadata in the original cifti file (palette settings, the command that generated the file and the commands that generated each of the inputs to that command).

This might explain the file size reduction.

Thanks,

Shilpi