Use MifImage class instead of mrconvert

[tsalo]

Closes #15. Closes #17.

Keeping as a draft for now.

I've opened nipy/nibabel#1489 to add the MifImage class to nibabel. Depending on how quickly that PR progresses, we might want to keep copies of the class in ModelArrayIO for a release or two.

NOTE: Since HBCD does not include fixel data, this is a low priority item.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 72.70115% with 95 lines in your changes missing coverage. Please review.
✅ Project coverage is 75.39%. Comparing base (cc9b983) to head (bb5bcd3).

Files with missing lines	Patch %	Lines
src/modelarrayio/utils/mif_image.py	71.91%	43 Missing and 23 partials ⚠️
src/modelarrayio/utils/_mif_format.py	73.17%	13 Missing and 9 partials ⚠️
src/modelarrayio/cli/h5_to_mif.py	72.22%	4 Missing and 1 partial ⚠️
src/modelarrayio/utils/mif.py	84.61%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #43      +/-   ##
==========================================
+ Coverage   70.35%   75.39%   +5.04%     
==========================================
  Files          19       21       +2     
  Lines        1167     1463     +296     
  Branches      228      304      +76     
==========================================
+ Hits          821     1103     +282     
+ Misses        284      266      -18     
- Partials       62       94      +32     

☔ 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.

[tien-tong]

will the template ever be scaled? Codex review said this, not sure if accurate: scaled template MIFs will produce numerically wrong outputs. src/modelarrayio/cli/h5_to_mif.py (line 68) copies the template header verbatim, while src/modelarrayio/utils/mif.py (line 582) has already applied header scaling to the in-memory data and the writer (line 395) writes the same scaling metadata back out. Any non-default scaling in example_mif will be applied twice.

[tsalo]

I know very little about the MIF format. @mattcieslak do you know the answer to this?

[tien-tong]

src/modelarrayio/utils/mif.py is now doing three jobs at once: format parsing, nibabel image behavior, and cohort/fixel helpers. Splitting that module would make future review and testing much easier. The duplicated result-writing blocks in h5_to_mif.py (line 62) could also be factored into one helper.

Documentation: docs are stale. docs/installation.rst (line 17) still says MRtrix mrconvert is required, and docs/examples/plot_fixel_workflow.py (line 133) still says existing files are not overwritten.

[tsalo]

@tien-tong I think I've addressed all of your comments except for the one on scaling.

[tien-tong]

Looks good to me!

[mattcieslak]

We should actually test this with some real fixel data, but I think it's pretty solid