Some initial processing of images can help make things work more reliably and accurately in FSL.
- All modalities:
Convert from DICOM (using your preferred tool: e.g. mricron/dicom2nii, freesurfer/mri_convert, etc.)
- Check that the conversions are good and that there are no serious artefacts present (open the images in FSLView and spend a couple of minutes looking at them - use the movie loop for functional and diffusion images)
fslreorient2std (assists with future registration steps and makes the views in FSLView more "standard")
- Structural images (in addition to the above):
cropping to remove neck and lower head: robustfov (automatic) or fslroi (manual)
- acquisition note: for T1w structurals it is very helpful to have some fat saturation in the T1w sequences to avoid marrow/meninges being mistaken for brain tissue in segmentation and registration