ENH: Add number of volumes per *b*-value to QC report

[josephmje]

Closes #64 and #128. Replaces PR #73.

• report diffusion sampling scheme: shelled, radial or Cartesian
• report shell distribution
• expose bmag parameter to adjust rounding bvals
• check # b0s doesn't change after rounding
• create tests for DiffusionSummary interface

In the case of DSI data, should round_bvals automatically round to the nearest 10 instead of calculating based on the magnitude of bvals? Set bmag = 1.

[oesteban]

• Why don't we also print out the shell distribution?
• What happens with DSI schemes? Maybe we can add a unit test of the DiffusionSummary interface?

[josephmje]

Right now, shells_dist does print out the shell distribution. But perhaps I should give it a different label and variable name to make this more clear.

I've tested this on the taiwan_ntu_dsi dataset that comes bundled with dipy. Here's what it outputs:

{0: 7, 100: 12, 200: 8, 400: 6, 600: 24, 900: 24, 1500: 12, 1900: 30, 2400: 24, 2900: 24, 3400: 8, 4000: 24}

I haven't worked with DSI data before so I'm not quite sure if this is what we're looking for or whether this should be improved.

Will update with a unit test for the new interface!

[oesteban]

Could you rebase when #151 is merged?

[josephmje]

I'll update this to see if I can figure out a parser that determines whether the data is shelled or sampled along the Cartesian grid. I've checked out the methods section of a few DSI papers and they seem to either report the maximum b value or a range (eg. x DWIs with diffusion weightings in the range of b = 100-4000 s/mm2 and x interleaved b = 0 scans). Perhaps a similar approach can be taken in our reportlet.

[oesteban]

I'll update this to see if I can figure out a parser that determines whether the data is shelled or sampled along the Cartesian grid. I've checked out the methods section of a few DSI papers and they seem to either report the maximum b value or a range (eg. x DWIs with diffusion weightings in the range of b = 100-4000 s/mm2 and x interleaved b = 0 scans). Perhaps a similar approach can be taken in our reportlet.

Yes, this sounds fantastic.

[pep8speaks]

Hello @josephmje, Thank you for updating!

Cheers! There are no style issues detected in this Pull Request. 🍻 To test for issues locally, pip install flake8 and then run flake8 dmriprep.

Comment last updated at 2021-03-11 15:15:14 UTC

[josephmje]

This is ready for review. I will leave the DSI updates for another PR.

[josephmje]

@oesteban tests are failing after some of the changes to the dwi reference workflow. will wait until #153 gets merged and then will rebase

[oesteban]

sounds great

[oesteban]

should we resuscitate this?