Quantitative registration QC and maybe even automatic optimization

[TheChymera]

Ideally we would like to have a function which can check the registration quality at a glance - to quantitatively compare alternative preprocessing workflows to ours, but also different parameter specifications for our own workflows. This feature could also be used to automate parameter optimization for our registration. This may require quite a bit of work, but it would be a splendid occasion to implement some machine learning for neuroimaging (we have enough data for this) and may actually be unavoidable if manual optimization remains fruitless.

More details and a bit of backstory in a conversation on the ANTs repo.

[TheChymera]

We start by making a similarity metric script from ANTs accessible via nipype.

[TheChymera]

Registration QC functions are available via the IRSABI manuscript:

https://www.biorxiv.org/content/10.1101/619650v3

Session-wise qualitative inspection will shortly be integrated from the work done on the NVCZ article: https://gitlab.com/Chymera/nvcz/-/blob/91000392dc40f6225ab7506d420f44c64ba4731b/prepare/preprocess_qc.py and multi-session QC should be integrated from the aforementioned article code: https://bitbucket.org/TheChymera/irsabi/src/c151cbc67007e73cbae6b9f9f08c2fa24b92afee/prepare/manual_overview.py#lines-136