C-PAC: Pipelines

Pre-Configured Pipelines: Default

C-PAC runs this pipeline by default, it is not necessary to invoke the –preconfig flag to run it.
The default processing pipeline performs fMRI processing using four strategies, with/without global signal regression, and with/without bandpass filtering.

Anatomical processing begins with conforming the data to RPI orientation and removing orientation header information that will interfere with further processing.
A non-linear transform between skull-on images and a 2mm MNI brain-only template are calculated using ANTs.
Images are then skull-stripped using FSL’s BET and subsequently segmented into WM, GM, and CSF using FSL’s FAST tool.
The resulting WM mask is multiplied by a WM prior map that has been transformed into individual space using the inverse of the linear transforms previously calculated during the ANTs procedure.
A CSF mask is then multiplied by a ventricle map derived from the Harvard-Oxford atlas distributed with FSL.
Skull-stripped images and grey matter tissue maps are written into MNI space at 2mm resolution.

Functional preprocessing begins with resampling the data to RPI orientation and performing slice timing correction.
Next, motion correction is performed using a two-stage approach in which the images are first coregistered to the mean fMRI and then a new mean is calculated and used as the target for a second coregistration (AFNI 3dvolreg).
A 7 degree of freedom linear transform between the mean fMRI and the structural image is calculated using FSL’s implementation of boundary-based registration.
Nuisance variable regression (NVR) is performed on motion corrected data using a 2nd order polynomial, a 24-regressor model of motion[7], 5 nuisance signals, identified via principal components analysis of signals obtained from white matter (CompCor[8]), and mean CSF signal.
WM and CSF signals are extracted using the previously described masks after transforming the fMRI data to match them in 2mm space using the inverse of the linear fMRI-sMRI transform.
The NVR procedure is performed twice, with and without the inclusion of the global signal as a nuisance regressor.
The residuals of the NVR procedure are processed with and without bandpass filtering (0.01Hz < f < 0.1Hz), written into MNI space at 3mm resolution and subsequently smoothed using a 6mm FWHM kernel.

Several different individual level analysis are performed on the fMRI data including: