ENH: Add writer for Siemens CSA header

[moloney]

Allows us to take a parsed CSA header and convert it back into a
string. Useful for things like DICOM anonymization, or perhaps
round tripping DICOM -> Nifti -> DICOM.

[moloney]

This is another bite-sized chunk from #290.

I was tempted to make a csa module with the read/writer functionality and then setup some proxies in csareader for backwards compatibility, but I am not sure it is worth the hassle?

It would also be nice to use ElemDict objects (as specified in #416) for the parsed representation. Again, I guess we need to provide some backwards compatibility here.

[nibotmi]

☔ The latest upstream changes (presumably #393) made this pull request unmergeable. Please resolve the merge conflicts.

[matthew-brett]

The csa representation as we wrote it originally is a bit kludgy, I think - if you're interested - it would be worth improving that with a new csa module, and depracting the csareaders module.

I don't think we have to worry as much as usual about backwards compatibility - the nicom module has always had a bit fat warning on it. So, avoid breakage if possible, but don't do excessive work to keep compatibility otherwise.

[matthew-brett]

By the way - thanks very much for breaking out the change from the big PR - it's extremely helpful.

[codecov]

Codecov Report

Merging #417 into master will decrease coverage by 0.03%.
The diff coverage is 81.63%.

@@            Coverage Diff             @@
##           master     #417      +/-   ##
==========================================
- Coverage   91.70%   91.66%   -0.04%     
==========================================
  Files          96       96              
  Lines       12311    12360      +49     
  Branches     2173     2186      +13     
==========================================
+ Hits        11290    11330      +40     
- Misses        684      690       +6     
- Partials      337      340       +3     

Impacted Files	Coverage Δ
nibabel/nicom/csareader.py	85.86% <81.63%> (-1.54%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 65d5fc6...122a923. Read the comment docs.

[moloney]

@ZviBaratz @matthew-brett @effigies

During our recent meeting, I forgot the original reason I wrote this was to support anonymization of Siemens mosaic/DWI Dicom files (not round-tripping back to DICOM). The standard way to anonymize DICOM removes all private headers, but then you lose the mosaic info and b-values/b-vecs. You can whitelist the CSA sub-header with this info, but at least some versions of the scanner software will use uninitialized memory for the CSA buffer and then leak PHI in the unused portions. So by parsing it and rewriting it with this code (which does zero out the memory) you can avoid this issue. So... it is pretty niche but still useful.

This also got me thinking, is dicom_parser going to make any attempt at supporting anonymization? Should we start a dicom_anon project?

[ZviBaratz]

I haven't really given it much thought so far, but I definitely support the idea. Do you think a separate package will be better? I feel like it could probably be added to dicom_parser fairly easily, but it's possible I'm simply not aware of some anonymization edge-cases, such as the CSA thing.