### Post-processing of fmriprep outputs

The eXtensible Connectivity Pipeline- DCAN (XCP-D)
[@mehta2024xcp;@mitigating_2018;@satterthwaite_2013]
was used to post-process the outputs of *fMRIPrep* version 25.2.4
[@esteban2019fmriprep;@esteban2020analysis, RRID:SCR_016216].
XCP-D was built with *Nipype* version 1.10.0 [@nipype1, RRID:SCR_002502].

#### Segmentations

The following atlases were used in the workflow: the Schaefer Supplemented with Subcortical Structures (4S) atlas [@Schaefer_2017;@pauli2018high;@king2019functional;@najdenovska2018vivo;@glasser2013minimal] at 10 different resolutions (1056, 156, 256, 356, 456, 556, 656, 756, 856, and 956 parcels), the Glasser atlas [@Glasser_2016], the Gordon atlas [@Gordon_2014], the Tian subcortical atlas [@tian2020topographic], and the HCP CIFTI subcortical atlas [@glasser2013minimal].
 Each atlas was warped to the same space and resolution as the BOLD file.

#### Anatomical data

Native-space T1w images were transformed to MNI152NLin2009cAsym space at 1 mm3
resolution.


#### Functional data

For each of the one BOLD runs found per subject (across all tasks and sessions),
the following post-processing was performed.

 The first four volumes of both the BOLD data and nuisance regressors were discarded as non-steady-state volumes, or 'dummy scans'.   In total, 36 nuisance regressors were selected from the preprocessing confounds,
according to the '36P' strategy.
These nuisance regressors included
six motion parameters, mean global signal, mean white matter signal,
mean cerebrospinal fluid signal with their temporal derivatives,
and quadratic expansion of six motion parameters, tissue signals and
their temporal derivatives [@benchmarkp;@satterthwaite_2013].

The BOLD data were despiked with *AFNI*'s *3dDespike*.

Nuisance regressors were regressed from the BOLD data using a denoising method based on *Nilearn*'s
approach.
The timeseries were band-pass filtered using a(n) second-order Butterworth filter, in order to retain signals between 0.01-0.08 Hz. The same filter was applied to the confounds. The resulting time series were then denoised using linear regression.  The denoised BOLD was smoothed using *Nilearn* with a Gaussian kernel (FWHM=4.0 mm).

Processed functional timeseries were extracted from the residual BOLD signal
with *Nilearn's* *NiftiLabelsMasker* for the atlases.
Corresponding pair-wise functional connectivity between all regions was computed for each atlas,
which was operationalized as the Pearson's correlation of each parcel's unsmoothed timeseries.
In cases of partial coverage, uncovered voxels (values of all zeros or NaNs) were either
ignored (when the parcel had >0.0% coverage)
or were set to zero (when the parcel had <0.0% coverage).


Many internal operations of *XCP-D* use
*AFNI* [@cox1996afni;@cox1997software],
*ANTS* [@avants2009advanced],
*TemplateFlow* version 25.1.1 [@ciric2022templateflow],
*matplotlib* version 3.10.5 [@hunter2007matplotlib],
*Nibabel* version 5.3.2 [@brett_matthew_2022_6658382],
*Nilearn* version 0.13.0 [@abraham2014machine],
*numpy* version 2.2.6 [@harris2020array],
*pybids* version 0.21.0 [@yarkoni2019pybids],
and *scipy* version 1.15.3 [@2020SciPy-NMeth].
For more details, see the *XCP-D* website (https://xcp-d.readthedocs.io).


#### Copyright Waiver

The above methods description text was automatically generated by *XCP-D*
with the express intention that users should copy and paste this
text into their manuscripts *unchanged*.
It is released under the [CC0](https://creativecommons.org/publicdomain/zero/1.0/) license.

#### References