[WIP] - MTMS-CSD Tutorial

[ShreyasFadnavis]

This tutorial is also intended to delineate DIPY's ability to create pipelines.
Steps involved:

• Denoising
• Registration
• Tissue Classifier
• MCSD

Addresses #1870

[pep8speaks]

Hello @ShreyasFadnavis, Thank you for submitting the Pull Request !

• In the file doc/examples/reconst_mcsd.py:

Line 44:1: E265 block comment should start with '# '
Line 46:1: E265 block comment should start with '# '
Line 48:1: E265 block comment should start with '# '
Line 67:1: E265 block comment should start with '# '
Line 81:1: E265 block comment should start with '# '
Line 82:1: E265 block comment should start with '# '
Line 85:1: E265 block comment should start with '# '
Line 124:1: E265 block comment should start with '# '
Line 125:1: E265 block comment should start with '# '

Do see the DIPY coding Style guideline

[codecov-io]

Codecov Report

Merging #1909 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #1909   +/-   ##
=======================================
  Coverage   85.43%   85.43%           
=======================================
  Files         119      119           
  Lines       14298    14298           
  Branches     2243     2243           
=======================================
  Hits        12215    12215           
  Misses       1575     1575           
  Partials      508      508

[ShreyasFadnavis]

Hi all! I am facing some issues with the data for this example.. The only viable option that I see from the fetcher is CFIN, but needs some brain extraction on the T1.
I have a working example with the HCP data, but we do not have a fetcher for it! any suggestions would be super helpful..

[jhlegarreta]

The HCP data requires a username/pwd to be able to download it. @arokem mentioned in PR #1900 that the terms and conditions of the HCP data prevent to redistribute them, and this is one of the reasons for having the CENIR dataset in DIPY. CENIR does not have T1-weighted images, though.

I had a quick look at the UW-Minn HCP data terms, and according to point 4, data and derivatives may be re-distributed under the same data use terms. So for that part of the data may be DIPY could comply with those restrictions? May be confirmation should be asked for to UW-Minn HCP team?

If that is feasible and the data of interest for DIPY is part of the UW-Minn HCP data, then may be it could be hosted in UW's digital library, as it is the case of the CENIR data, and the corresponding fetcher could then be built?

[ShreyasFadnavis]

The HCP data requires a username/pwd to be able to download it. @arokem mentioned in PR #1900 that the terms and conditions of the HCP data prevent to redistribute them, and this is one of the reasons for having the CENIR dataset in DIPY. CENIR does not have T1-weighted images, though.

I had a quick look at the UW-Minn HCP data terms, and according to point 4, data and derivatives may be re-distributed under the same data use terms. So for that part of the data may be DIPY could comply with those restrictions? May be confirmation should be asked for to UW-Minn HCP team?

If that is feasible and the data of interest for DIPY is part of the UW-Minn HCP data, then may be it could be hosted in UW's digital library, as it is the case of the CENIR data, and the corresponding fetcher could then be built?

Hi @jhlegarreta ! I completely agree with you. I have reached out to the HCP team and am awaiting a response. In the meantime, let me also take a look at the clauses of UW-Minn HCP ✔

[arokem]

Sorry - I believe that the HCP data is a no go for us. If I understand correctly, you can redistribute the data, but you have to put in place a mechanism for downstream users to also agree to the original terms of conditions, and we can't do that in our software. Please use the CENIR data, or some other data. Both here and in #1900.

[ShreyasFadnavis]

Hi @arokem , the issue is that CENIR does not have a T1. The next option that we have using CFIN, but is not extracted, will need something like BET + registration to do this.. Can do this, but I dont want to go outside DIPY tools for the tutorial. Does this make sense?

[jhlegarreta]

@skoudoro Sorry to chime in the middle of the release process, but I am wondering why the file added in this example is not shown in the master branch (either in GitHub or when fetching the branch locally). Am I missing something?

BTW, was the example left out of the examples_index.rst on purpose (e.g. needs polishing)?

[skoudoro]

Strange...

• First, This tutorial should not be merged it is not ready at all...
• Second, I do not remember that I merged this PR, it is not even on my mailbox,
• Third, I do not see it on master too...

Something weird happens.... Github bug? need to investigate....

Thank you for pinging me @jhlegarreta !

[skoudoro]

I really do not know..... Somehow, this PR is linked to #1931 which I merged yesterday. I suppose #1931 was build on the top of this branch but that's still weird, I really do not know what happens here! @arokem @Garyfallidis, @ShreyasFadnavis any idea?

[arokem]

Looks like it wasn't actually merged at all, right? So no big deal (though mysterious...). I think that @ShreyasFadnavis would need to create a new PR. Either from this branch or from a new branch.

[ShreyasFadnavis]

@jhlegarreta Thank your for pointing this out! This is scary! @arokem @skoudoro are right that this was never merged is probably a bug! Will open up a new PR for this. Nevertheless, any idea on what data set to use for this example? This tutorial needs T1 (extracted) + DWI...

[arokem]

Would a "pseudo-T1" work? You can create one from DWI data with an Anisotropic Power Map (see e.g., https://gist.github.com/arokem/5d7441902669a4a82389)

[ShreyasFadnavis]

Thanks @arokem !! Let me this a shot... it should work I guess ✔