Normalization to a custom common template

fmriprep
fmri

#1

Hi,

I was using normalization (MNI template) till now, and wandering whether to experiment with other templates.

  1. How does one extracts common template from fmriprep’s output?
  2. What are the parcellation options when normalizing to a such a common template created from the data?
    (nilearn extract regions? anything else?)
  3. In case of new data- it would be nice to see an example of bringing the test, new data, to the template calculated from the training data (I understand what should be done, but a reference to an example would be nice)
  4. Anyone recommends applying the method described here (or- we have anatomical data- could it be that we better not use it???):
    https://hal.archives-ouvertes.fr/hal-01701619/file/frontiers.pdf
    Any implementations to this method in the open source libraries?

Many many thanks!


#2

From what I hear it is very time consuming (computationaly) with diminishing returns (unless you are using a very unusual population, such as kids).

You should have a look at https://github.com/ANTsX/ANTs/blob/master/Scripts/buildtemplateparallel.sh

One thing I would explore is calculating transformation (even affine) between your custom template and the MNI. This way you gain (slightly) better alignment across participants and can still use parcels defined in the common space.

A variation of this approach (without custom template building though) is on our TODO list https://github.com/poldracklab/fmriprep/issues/620. If you care about this feature, we would love your help adding it.


#3

LOVE your answers.
Thanks a lot.
(Will update if I find the time to help with this task)