Update fMRIPrep-related documentation

[tsalo]

Closes none. I'm trying to translate the settings we're planning to use in fMRIPost-rapidtide into recommendations in rapidtide's documentation.

Changes proposed in this pull request:

• fMRIPrep outputs all of the transforms you might need, so I removed a comment about extracting transforms from the working directory.
• Change recommended interpolation method for aparc+aseg files from NearestNeighbor to GenericLabel.
• I added a tip to use the gray matter mask derived from fMRIPrep for --globalmeaninclude (already recommended), --refineinclude, and --offsetinclude.
  • @bbfrederick I want to make sure this is a good tip- it's what I have in fMRIPost-rapidtide, so if I'm wrong to recommend it I should modify fMRIPost-rapidtide as well.
• Some formatting changes to the associated sections. I broke a few lines on punctuation and code-formatted arguments in the text.

[bbfrederick]

All good suggestions and changes. The wording about extracting transforms from the working directory is from the bad old days before these were output by default.

I'm leaning towards using the gray matter mask for globalmeaninclude as well. In fact, along those lines, there is a new capability in the last few versions that I haven't documented yet - instead of supplying the masks individually, you can just give rapitdide --brainmask, --graymattermask and --whitemattermask, and it sets the internal masks based on whatever current theory I have about what regions are good to use for each one (if you explicitly set any of the masks on the command line it overrides the "standard" behavior for that mask.

[bbfrederick]

I'm not sure what's making build_docs fail - sphinx is a bit of a black box to me; the documentation broke a while back and I threw code at the wall until it started working again then ran way - I'm afraid to touch it now :-).

[tsalo]

I have been using those three parameters when processing 28andMe. I'm happy to modify the fMRIPrep recommendations to use those parameters instead. If you use --graymattermask, should you still use the same mask for --refineinclude as well? For healthy young adults, I mean.

Also... since these mask files are binarized automatically with a threshold of 0.1, do you still recommend binarizing fMRIPrep probseg files before using them?

The two problems I stumbled across are that the intersphinx mapping dictionary was formatted in a way that isn't supported anymore and m2r recently started breaking. m2r has been unmaintained for a while, so I believe the current recommendation is to use myst_parser, which is what I'm trying.

[bbfrederick]

I don't have any firm thoughts about what to use for --refineinclude. On the one hand, gray matter is likely to have the strongest signal, so you might get a higher quality regressor, but including white matter makes it more likely that you'll get a regressor that's more "balanced" to all tissue types (then again, if it should be the same regressor everywhere, the gray matter signal should be sufficient). For the time being, I think I'll leave --graymattermask as is - it will set --globalmeaninclude and --offsetinclude but not do anything to --refineinclude, unless you feel strongly that it should set --refineinclude too.

[tsalo]

I definitely don't have strong feelings about it. For the fMRIPrep recommendations, I can just not make one regarding refineinclude.

[tsalo]

Okay, I think the recommendations are fairly solid/up-to-date now. What do you think?

[bbfrederick]

Looks good. I'll merge it.

[bbfrederick]

Actually, NOW I'll merge it.

[tsalo]

Thanks! Although it looks like you closed without merging.

[bbfrederick]

That explains a lot. But why is build_docs failing again?

[tsalo]

I think that's just a dumb mistake on my part. I just pushed a fix, but you should wait to make sure it works before merging.

[tsalo]

Argh, the changelog page is being rendered (i.e., the URL exists), but it's not showing up in the sidebar.

[bbfrederick]

build_docs passed, and I have no reason to think the coverage test will fail (since no code changed). Merge?

[tsalo]

I just need to fix the changelog. I think I know what's wrong though.

[bbfrederick]

Ok. Are we .md free going forward?

[tsalo]

Not quite. It's actually basically the same as before, where whatsnew.rst just collects the contents of CHANGELOG.md, but using myst_parser instead of m2r or recommonmark, since both of the latter are unmaintained at this point.

[tsalo]

The sidebar looks good now!

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 34.56%. Comparing base (c598b78) to head (4058e26). Report is 34 commits behind head on main.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #172 +/- ## ========================================== + Coverage 34.45% 34.56% +0.10% ========================================== Files 135 135 Lines 24334 24334 Branches 3725 3501 -224 ========================================== + Hits 8385 8411 +26 + Misses 15129 15128 -1 + Partials 820 795 -25 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.