As everyone who read through the fmriprep documentation should know, the default MNI template used for normalization is MNI2009cAsym and not the MNI6 used by FSL or SPM.
This is posing a conundrum when you want to use resources that are MNI coordinate based from pretty much any other resource. Be it atlases, parcellations, ROI masks, meta-analyses etc’ that were made available by our friends and colleagues. This is a bit of an impediment for a project that is trying to promote collaboration and standardization across the field.
The two solutions that I was able to come with in my research are:
- Use MNI6 output and forego the advantages that led to the choice of MNI2009 as the default space.
- Track any other resource that I am interested in and transform it using the available transformation from https://figshare.com/articles/MNI_T1_6thGen_NLIN_to_MNI_2009b_NLIN_ANTs_transform/3502238 , which as was discussed in a different topic, is not ideal because there are version differences even within the SPM/FSL world, and the transform is to 2009b not 2009c that used by fmriprep. However, those differences seem minimal compared to the differences between the 6th gen and 2009.
So currently I mainly do option 2 and I have a local ever-growing library of fmriprep compatible resources which feels very idiosyncratic to my research interest.
I guess that I am starting this topic with 3 goals:
- Raise awareness to the issue of output spaces to any fmriprep newcomer that stumbles across the post and have not thought about the fact that looking at coordinates in MNI space is not quite enough.
- Ask if I am missing something or if there are any other recommendations from other users.
- See if there’s any interest in creating something of a repository for the fmriprep users community where we can share resources that will be ready to use (barring voxelsize resampling) with our fmriprep preprocessed data. And if so, how to go on about it.