Get XCP-D compatible with MNIPediatricAsym spaces

[KirkGraff]

Hi XCP-D devs, I think it would be really helpful if XCP-D was compatible with pediatric templates. It's currently functional with infant and adult data, so it seems like other age ranges is the obvious next step. Personally, I'm biased because I've been using the MNIPediatricAsym template (4.5-8.5 y.o), so that would be my preference.

Alternatively, if anyone has any suggestions for getting data that has been run through fmriprep, outputing to non-standard spaces, to work with XCP-D, that would be appreciated.

Thanks!

[tsalo]

The main blocker for supporting new atlases (at least for the NIfTI workflow) is that we need transforms to and from MNI152NLin6Asym, since that's the space XCP-D's atlases are in. Unfortunately, I don't see any transforms from the MNIPediatricAsym atlases to MNI152NLin6Asym on TemplateFlow. If you can find those transforms, I'd be happy to take a crack at supporting MNIPediatricAsym.

I'm comfortable including transform files in the XCP-D repository temporarily, with the caveat that they will need to be added to TemplateFlow at some point in the future.

[KirkGraff]

Thanks for the reply. Is there anything special about these transforms? Could I just use ANTS to register PediatricAsym to MNI152NLin6Asym, generating a .h5 file? Presumably a "more official" transformation exists elsewhere - though perhaps PediatricAsym<->MNI152NLin2009cAsym might be easier to find.

It'd be useful if XCP-D was setup to be usable in any space, provided the user provides a transformation to MNI152NLin6Asym space, though that might be beyond the scope of the goals of XCP-D.

[tsalo]

Is there anything special about these transforms? Could I just use ANTS to register PediatricAsym to MNI152NLin6Asym, generating a .h5 file?

To build new transforms between templates for templateflow, there's actually a whole process documented in https://github.com/templateflow/registration-framework. We (@mattcieslak and I) have run that workflow to generate transforms between MNI152NLin6Asym and MNI152NLin6Sym in the past. It requires data that we received from contacting Oscar Esteban.

However, I don't know if the same process would work well between templates from very different cohorts (e.g., across age ranges). If these transforms aren't already available, it would be a good idea to contact the developers of the pediatric templates to see if they have created them internally.

perhaps PediatricAsym<->MNI152NLin2009cAsym might be easier to find.

If you can find that transform, that would be fine. XCP-D can chain transforms together, and we already have MNI152NLin6Asym <-> MNI152NLin2009cAsym working.

It'd be useful if XCP-D was setup to be usable in any space, provided the user provides a transformation to MNI152NLin6Asym space, though that might be beyond the scope of the goals of XCP-D.

Yes, this is outside of XCP-D's scope. Sorry about that.