## Major changes from fMRIPrep
### easy stuff
- Default template is Fisher344
- Default structural is T2w
- No surface processing
- ~~No ICA aroma~~
- CSF mask eroded by 1mm
- segmentation (FAST) should not run any bias field correction
- FSL commands (FAST/~~FLIRT~~) must have `--basescale 1` to prevent undesired upscaling
- ANTS is used instead of FLIRT for coregistration
- BOLD resampling should use native resolution, not template

### less easy stuff
- brain extraction uses `nirodents`
- in functional reference workflow (create one reference image per BOLD run)
- structural workflow
- `3dVolreg` used for moco, rather than `MCFLIRT`
- no RMSD
- `N4BiasFieldCorrection`
- voxel sizes are (often) anisotropic
- requires a bspline grid rather than single distance value
- reduced shrink factors (4->1) minimise the chance of blurring over out-of-plane dimension
- SDC:
- if PE dir is IS/SI need to reorient to LSA explicitly before passing to topup
- need to scale image header so minimum voxel size is 1mm (for compatibility with topup acqparams file)
- -> outputs are then returned to original scale
- No *a priori* field map (yet) - but this might not be needed

## Goals outstanding
- Bump dependencies to match fmriprep
- merge SDC workflow PR ([#55](https://github.com/nipreps/fmriprep-rodents/pull/55))

Is it worth trying to complete these, or should we just take "the spirit" of the commits and implement into fMRIPrep?

## Infant changes
- Level of processing:
- Session
- Average
- T2w reference anatomical
- Tissue segmentation
- Joint Label Fusion
- Age parsing, template matching
- Alternative surface reconstruction
- MCRIBS
- Infant FS
- Multi-step normalization
- Anatomical -> MNIInfant -> MNI152
-