ENH: Add support for other petref images for registration

[mnoergaard]

This PR addresses issue #184 and #177 by adding support for 1) using other reference PET images for e.g. co-registration, and 2) implementing an additional option for performing PET to anatomical co-registration using ANTs. It implements the possibility to use 'twa' which is a time-weighted average of the motion corrected PET data. Furthermore, it also allows the use of 'sum' which is a summed image across time for the motion corrected PET data. The default is still the template generated from the motion correction using mri_robust_template.

[codecov]

Codecov Report

❌ Patch coverage is 86.17886% with 34 lines in your changes missing coverage. Please review.
✅ Project coverage is 80.81%. Comparing base (133526d) to head (03c6a78).

Files with missing lines	Patch %	Lines
petprep/workflows/pet/fit.py	79.13%	11 Missing and 13 partials ⚠️
petprep/interfaces/reports.py	60.00%	3 Missing and 3 partials ⚠️
petprep/cli/parser.py	33.33%	1 Missing and 1 partial ⚠️
petprep/workflows/pet/registration.py	95.45%	1 Missing ⚠️
petprep/workflows/pet/tests/test_fit.py	98.83%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #185      +/-   ##
==========================================
+ Coverage   80.63%   80.81%   +0.17%     
==========================================
  Files          84       84              
  Lines        6512     6738     +226     
  Branches      657      689      +32     
==========================================
+ Hits         5251     5445     +194     
- Misses       1097     1112      +15     
- Partials      164      181      +17     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[mnoergaard]

@mathesong - think this is ready for review. Will continue to do more testing locally on my end and test robustness/performance. We may want to consider making ants the default option for co-registration. WDYT? But this should go into a separate PR.

[mathesong]

Looks good to me!

I've been testing this over the weekend on my other dataset too, and unfortunately it's not a silver bullet. I still have failures in ~25%. Some overlapping from before, some not. So better than before, but not quite as robust as I had hoped. I then ran it with the hmc with a 2mm fwhm, and it's saved a few, but they're still causing issues. So I'm not really sure what the default ought to be if we still have issues :\

I suspect that the primary issue is that for this tracer, there's just so much uptake outside the brain in the PET, which makes things messy. There's got to be a solution somewhere...

[mathesong]

I just started running a few of the problematic subjects with the updated ANTS settings from ~30 minutes ago. I see you reverted some of them? Should I stop and restart with the new settings?

Also, another thought is that it might be worth having another pet2anat method called "ants_withskull" where the brain mask is dilated by 2cm around the brain, and then masked with the head mask from FreeSurfer. This might just give a bit more information for cases where there's either a lot going on around the brain, or alternatively where there's just not so much binding in the brain like when there's high occupancy.

[mnoergaard]

I just started running a few of the problematic subjects with the updated ANTS settings from ~30 minutes ago. I see you reverted some of them? Should I stop and restart with the new settings?

Also, another thought is that it might be worth having another pet2anat method called "ants_withskull" where the brain mask is dilated by 2cm around the brain, and then masked with the head mask from FreeSurfer. This might just give a bit more information for cases where there's either a lot going on around the brain, or alternatively where there's just not so much binding in the brain like when there's high occupancy.

The updated settings failed for another dataset I have, where there is not much uptake within the brain. So reverted my changes. Interestingly, for two of the ones that failed on my end with ANTs, worked with mri_coreg using --petref sum. This makes we think that we may also want to support more options for generating the PET reference image, such as a summing over the first 5 minutes to create a flow-like image. WDYT? Nevertheless, more testing is needed, and I think it is good to allow for several additional options, as some settings will be good for some tracers, and other settings will be good for other tracers.

It's a good idea with ants-with-skull!