add other coil compressions

[aTrotier]

I am preparing the field to add other coil compressions. Are you ok with the modification before going further ?

[tknopp]

Are you ok with the modification before going further ?

yes, looks fine!

[codecov-commenter]

Codecov Report

Base: 63.05% // Head: 63.05% // No change to project coverage :thumbsup:

Coverage data is based on head (5d3fdee) compared to base (1ee55aa). Patch has no changes to coverable lines.

:exclamation: Current head 5d3fdee differs from pull request most recent head ecd26f4. Consider uploading reports for the commit ecd26f4 to get more accurate results

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #131 +/- ## ======================================= Coverage 63.05% 63.05% ======================================= Files 8 8 Lines 249 249 ======================================= Hits 157 157 Misses 92 92 ``` Help us with your feedback. Take ten seconds to tell us [how you rate us](https://about.codecov.io/nps?utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=MagneticResonanceImaging). Have a feature suggestion? [Share it here.](https://app.codecov.io/gh/feedback/?utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=MagneticResonanceImaging)

:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.

[aTrotier]

I have added the coil compression directly on kspace : kx,ky,kz,coils,echos,repetitions I think it will be good to put the slice dimension maybe along dimension 7.

Right now, kDataCart extracts the slice along the dim 3 but when I want to convert back the kspace in an acquisition object I assume it is a 3D kspace. same for coilcompression, I assume the kspace is in 3D and not a multislice acquisition.

[aTrotier]

I have added gcc when passing a kspace (kx,ky,kz,coils,echos,repetitions)

As suggested by @JeffFessler, @alexjaffray and @JakobAsslaender for fieldmaps/espirit, should we implement most of the method in order to work on the kspace and make an interface with AcquisitionData ? (at least for cartesian data)

I don't know if it is possible to apply gcc more directly without building a kspace like array in order to apply it along others dimensions.

[aTrotier]

Results on the datasets available here (32 channels) : https://people.eecs.berkeley.edu/~mlustig/software/ESPIRiT_demo/demo_GCC.html

• RMSE SCC (32 -> 6) : 8.0e-5
• RMSE GCC (32 -> 6) : 3.5e-5

[tknopp]

I can't review this in detail right now but the test look all good and since this is definitely a clear improvement I merge this now. Improvements can be made later on.