Two (hopefully) brief question about mitigating the effects of brain cropping Mitigating the effects of brain cropping — MRtrix 3.0 documentation) in fixel based analyses; specifically about step 2.Ignore other instructions elsewhere regarding brain masks.
- mtnormalise does not allow you to not supply a mask. Is it acceptable to input a mask of the upsampled DWI here or should another mask file be used?
- the help of population_template states that providing a mask_dir will significantly speed up registration step. Is it ok here to use the same mask as for mtnormalise to speed things up, should another maskfile be used or is it better not to supply a mask_dir (and just be patient)?
mtnormalise requires the mask to match the image grid so if your dMRI data are upsampled, the mask needs to match that. You can use
mrgrid mask.mif regrid -template odf.mif -interp linear mask_regrid.mif
to regrid the mask to the the voxel grid of the ODF image data.
population_template uses masks for initialization and registration. If you used the option
-nanmask, then areas outside the input mask do not contribute to the template during the aggregation step. In registration and other contexts in MRtrix, NaN are effectively ignored, so the (mean or median) aggregation in areas with NaNs is performed over fewer samples. Areas outside the FOV of images never contribute to aggregation or registration irrespective of the options provided.
I’d nearly always use masks in
population_template unless, for instance, if any masks segment (mostly) non-overlapping anatomical areas as that would throw off initialization and prevent coregistration. Use
-nanmask if masks cover all parts of interest but if there are areas outside the masks that should not contribute to the template (lesions, image artefacts, …).
thanks for the feedback, Max!