BF: Ensure bvec/bval files are only created for dwi output

[pvelasco]

Currently, .bvec/.bval files are copied into the output folder whenever they are present in the dcm2niix output. dcm2niix (v1.0.20201102) creates those files whenever it detects a diffusion sequence and the images are not derived (rordenlab/dcm2niix#352 (comment)). So, with the current heudiconv, if you extract an image collected with a diffusion sequence, but with b-value = 0 as a fmap (pepolar modality), it will copy the .bvec/.bval files to the fmap folder. However, this doesn't conform to BIDS format, which only allows for .bvec/.bval for data in the dwi folder. Consequently, BIDS-validator (v1.5.10) gives an error:

bids-validator@1.5.10

	1: [ERR] Files with such naming scheme are not part of BIDS specification. This error is most commonly caused by typos in file names that make them not BIDS compatible. Please consult the specification and make sure your files are named correctly. If this is not a file naming issue (for example when including files not yet covered by the BIDS specification) you should include a ".bidsignore" file in your dataset (see https://github.com/bids-standard/bids-validator#bidsignore for details). Please note that derived (processed) data should be placed in /derivatives folder and source data (such as DICOMS or behavioural logs in proprietary formats) should be placed in the /sourcedata folder. (code: 1 - NOT_INCLUDED)
		./sub-FuncDiff/fmap/sub-FuncDiff_acq-dwi_dir-AP_epi.bval
			Evidence: sub-FuncDiff_acq-dwi_dir-AP_epi.bval
		./sub-FuncDiff/fmap/sub-FuncDiff_acq-dwi_dir-AP_epi.bvec
			Evidence: sub-FuncDiff_acq-dwi_dir-AP_epi.bvec
		./sub-FuncDiff/fmap/sub-FuncDiff_acq-dwi_dir-PA_epi.bval
			Evidence: sub-FuncDiff_acq-dwi_dir-PA_epi.bval
		./sub-FuncDiff/fmap/sub-FuncDiff_acq-dwi_dir-PA_epi.bvec
			Evidence: sub-FuncDiff_acq-dwi_dir-PA_epi.bvec

This PR makes sure we only copy the .bvec/.bval if the destination folder for an image is dwi. It also adds a unit test, and the necessary dicom file and heuristics.

[pvelasco]

It fails CI because the version it installs from Neurodebian is still 1.0.20181125, while the new behavior seemed to have been introduced in version 1.0.20201102.
I'll have to investigate some more.

[yarikoptic]

I think it would be better to be safe than sorry and check if bval is really 0. If not - issue a warning and not remove. That should also then alert bids validator, and user would need to investigate why they have those files (and why it wasn't 0 bval)

[codecov]

Codecov Report

Merging #491 (5099d01) into master (863435e) will increase coverage by 0.34%.
The diff coverage is 91.66%.

@@            Coverage Diff             @@
##           master     #491      +/-   ##
==========================================
+ Coverage   76.45%   76.79%   +0.34%     
==========================================
  Files          39       40       +1     
  Lines        3045     3090      +45     
==========================================
+ Hits         2328     2373      +45     
  Misses        717      717              

Impacted Files	Coverage Δ
heudiconv/convert.py	87.13% <84.21%> (+0.63%)	⬆️
heudiconv/heuristics/test_b0dwi_for_fmap.py	92.30% <92.30%> (ø)
heudiconv/tests/test_convert.py	100.00% <100.00%> (ø)
heudiconv/utils.py	90.32% <0.00%> (+0.40%)	⬆️

---

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 863435e...5099d01. Read the comment docs.

[pvelasco]

I think it would be better to be safe than sorry and check if bval is really 0. If not - issue a warning and not remove.

That sounds sensible. I will work on it.

[pvelasco]

@yarikoptic,
I think 9746fe6 addresses your suggestion. I have also modified the unit test accordingly.

[pvelasco]

you saw usecases with non absolute 0 bvals where they still should be considered? just curious

No. I would not use a diffusion-weighted image (b non zero) for fmap estimation purposes, because you will always get lower SNR than the equivalent b-value = 0 image.
Note: Siemens uses all the gradients in the sequence to compute the real b-value. So even if you select b-value=0, you get a value of 5 in the header.

[yarikoptic]

Great, thank you @pvelasco !