[ENH] update_complex_name to handle real+imaginary complex data

[bpinsard]

Someone recently shared with me some GE Hyperband multi-echo data from a SIGNA UHP, where they exported mag + real + imaginary data (weird combination but why not), all 3 ends up in the same series. So that should ideally be handled in heudiconv within update_complex_name which only covers mag+phase case for now.

Though I hope it won't be the final export format for that project, it could make sense to cover that usecase in heudiconv.

I don't know if there are any open test data with real+imaginary data within the same series, @neurolabusc ?

[bpinsard]

Codespell get called in both Codespell and Linters? It fails in both on some files that has nothing to do with that PR.

[codecov]

Codecov Report

Attention: Patch coverage is 62.50000% with 6 lines in your changes missing coverage. Please review.

Project coverage is 82.07%. Comparing base (383525d) to head (152fc2d).

Files with missing lines	Patch %	Lines
heudiconv/dicoms.py	16.66%	5 Missing ⚠️
heudiconv/convert.py	90.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #761      +/-   ##
==========================================
- Coverage   82.13%   82.07%   -0.07%     
==========================================
  Files          42       42              
  Lines        4226     4234       +8     
==========================================
+ Hits         3471     3475       +4     
- Misses        755      759       +4     

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

[bpinsard]

This will require rordenlab/dcm2niix#826 to work, otherwise we might need to assume that data are "MAGNITUDE" by default in update_complex_name

[yarikoptic]

Codespell get called in both Codespell and Linters? It fails in both on some files that has nothing to do with that PR.

thanks! now there is

• Codespell tuneup: no dedicated workflow (part of tox -e lint), and fix few freshly detected typos #762

[yarikoptic]

I guess let's wait on dcm2niix decision but meanwhile provided some recommenations to changes/around... can't do more today -- better go do smth as dumb as fix a car

[bpinsard]

ImageType is a mess: just remembered that diffusion data on Siemens won't have that info in the ImageType, see CMRR-C2P/MB#305 for the CMRR sequence where they were able to fix that.

I wonder if dcm2niix should add a new ComplexImageComponent json tag rather than editing the ImageType to match what inconsistently exists on Siemens.