Suitable for use on infant data?

[effigies]

Hi all, we're working on an infant version of fMRIPrep (nibabies), and were wondering how much we can expect fmridenoise to work with the outputs without configuration, and how much adaptation we need? I see some MNI152[...] segmentations so that part would presumably need some modification, but I wanted to get your opinions on the overall prospects.

cc @ericfeczko @mgxd

[SiegfriedWagner]

I have no experience with infant data, maybe @kfinc or @kbonna could give more insight.
You mentioned parcellation files, they are retrieved using code from fmridenoise.parcellation.get_parcellation_file_path.
I think I could modify the code and allow adding new directories with parcellation paths as console args

[effigies]

Yes, to refine the questions:

1. Are there other places that make strong assumptions about the size of the brain?
2. Are the pieces that use the templates mandatory, or bypassable?
3. Are you open to the addition of new templates?

The ability to add a template at run-time is definitely useful during process development, but to make it a drop-in next step to preprocessing, we would either need to publish an infant version of fmridenoise (as we are with fMRIPrep/nibabies) or merge any necessary pieces back here.

[SiegfriedWagner]

1. I don't think so
2. Mandatory (Connectivity interface)
3. Yes, also idea with custom spaces from command line seems to me viable.

I would like to get some feedback from @kfinc or @kbonna before pushing for changes further.

[kfinc]

Hi! Besides loading brain parcellations with corresponding distance matrices, we do not make strong assumptions about the size of the brain. These templates are mandatory, but we were considering to customize template selection -- that would be great to add more templates.

[effigies]

Thanks for the responses! It does sound like adding infant templates here (probably in the context of #29?) would be the most productive way forward.

We're not quite at that stage, but we'll keep in touch. Feel free to close this issue if it should be folded into #29.