Add PNG microscopy data and files for testing dataframe

[mariehbourget]

Related to #5
This PR aims to add microscopy data to the data-testing repo.
It also covers the _sessions.tsv and _scans-tsv files and architecture needed for the testing of the new loader in ivadomed.

Currently, I have an issue with the data that was already in the repo (MRI anat), because it is used for testing in ivadomed and the path is harcoded in several tests and config files:

• Ideally, we would want to have different BIDS dataset folders inside the data-testing repo (e.g. anat, microscopy, eeg, dwi, etc).
• So I added a microscopy BIDS folders for PNG format in microscopy_png and updated the README accordingly.
• However, files related to MRI anat are still at the root of data-testing (all tests passed).
• If I move them into a "mri-anat" dataset folder for example, 48 tests are failing because of the paths.

I would like to go ahead and write a test for the new loader. I can do it even if the data stays like that, but it is not very clean and I think it would be best to update the MRI anat data as well (in another PR) @lrouhier @charleygros what do you think? Thanks!

[lrouhier]

I think You are right this is not very clean, and even though 48 seems like a big number I don't think it would take too much time to fix (because this would only be a path_related issue). I don't know about the order though, should we break everything now or wait for the JOSS review to be over (because we will do a release after the review right ?)

[mariehbourget]

I don't know about the order though, should we break everything now or wait for the JOSS review to be over (because we will do a release after the review right ?)

I think we should wait for the JOSS review, I'll write the tests with the current folder hierarchy of this branch and I'll fix it afterwards.

[charleygros]

I think we should wait for the JOSS review, I'll write the tests with the current folder hierarchy of this branch and I'll fix it afterwards.

Ideally we should wait for the JOSS review. But we don't really know when it will finish (only one reviewer left tho): I don't want to freeze your development for an uncertain period of time :/

[mariehbourget]

Ideally we should wait for the JOSS review. But we don't really know when it will finish (only one reviewer left tho): I don't want to freeze your development for an uncertain period of time :/

I wrote the tests for the function with the file architecture of this branch in PR #584 in ivadomed repo.
All previous tests are still passing on ivadomed.

Here is what I suggest:

1. Merge this PR
2. Create a new release for data-testing
3. Update the link in download_data in PR #584 in ivadomed here to point to the new data-testing release
4. Merge PR 584 in ivadomed (all tests should still passed)
5. Wait for JOSS review
6. Open a new PR here to update files architecture, and a new PR on ivadomed to fix paths in tests.

If we want to be hyper-cautious, we can wait for JOSS review before step 4. This won't freeze the development on my part.

[jcohenadad]

If we want to be hyper-cautious, we can wait for JOSS review before step 4. This won't freeze the development on my part.

I think it's fine. Worst case scenario: there is a glitch for a few hours, while one reviewer is looking at the repos, but they would easily understand it's a temporary issue that will be fixed a few hours later. If it glitches happen to Google it can happen to us 😉

[mariehbourget]

This PR is ready for review.
After merged, a new release must be done in order to update the download link in #584 (ivadomed).
@charleygros or @andreanne-lemay, would you like to review?

[andreanne-lemay]

Looks great, thank you!

[mariehbourget]

Thank you @andreanne-lemay and @charleygros, I'll merge and create the new release.