Interpolated output of XCP-D desc-interpolated_bold.nii.gz

Summary of what happened:

I was reading the documentation about the output produced by XCP-D and I would like to ask why it is written, and I quote: “The interpolated denoised BOLD files (desc-interpolated) should NOT be used for analyses.”

My interest is because I would like to use the interpolated, temporal filtered, regressed and smoothed version of the fMRI bold data as I need a non-censored version for my later analysis.

Command used (and if a helper script was used, a link to the helper script or the command generated):

/usr/local/miniconda/bin/xcp_d /fmriprep_der/ /out_dir/ participant --participant-label CC110033 --nproc 12 --omp-nthreads 12 --mem-gb 24 --input-type fmriprep --nuisance-regressors acompcor --smoothing 6 --fd-thresh 0.5 --min-time 175 --lower-bpf 0.001 --upper-bpf 0.0 --atlases Glasser Tian -w /work/ --write-graph --fs-license-file /fs_dir/fs_license.txt --resource-monitor



Environment (Docker, Singularity / Apptainer, custom installation):


Data formatted according to a validatable standard? Please provide the output of the validator:

This dataset appears to be BIDS compatible.
        Summary:                   Available Tasks:        Available Modalities: 
        6524 Files, 29.34GB        rest                    MRI                   
        652 - Subjects                                                           
        1 - Session           

Relevant log outputs (up to 20 lines):


Screenshots / relevant information:


It’s up to you whether to use the interpolated data or not, but I wouldn’t recommend including “fake” (i.e., interpolated) volumes in analyses. What specifically do you need the non-censored data for in your later analyses? If the later analyses include censoring, then it’s not an issue to use the interpolated data.

Alternatively, do you need to include the censoring step in your XCP-D run? You could just run XCP-D without an FD threshold. Despiking is a reasonable alternative.

Okay thanks, my problem is that I cannot use censored data for my later analysis as I need continuous time series.

I wanted to check that the only difference between desc-interpolated and desc-denoised is the censoring of the volumes above the fd threshold (and also that the difference between desc-interpolated and desc-denoisedSmoothed is simply for censoring and smoothing), could you assure me of this?

That is correct. Just be aware that desc-interpolated has interpolated (i.e., fake) data in those high-motion volumes.

Great, thanks a lot for the help, very kind of you.