Error creating gradient_matrix.txt in DTIRecon (Diffusion Toolkit)

[mmromero]

Summary

When the gradient_matrix.txt file is created from bvecs, the lines are read in order y, z, x; instead of x, y, z swapping all the directions in the resulting gradient_matrix.txt file

Actual behavior

The columns of the actual gradient_matrix.txt are y, z, x, b-value.

Expected behavior

The columns of the expected gradient_matrix.txt must be x, y ,z, b-value.

How to replicate the behavior

Just run the example in the documentation of nipype interface for Diffusion Toolkit. My data comes from a Philips Scanner.

Platform details:

{'scipy_version': '1.0.0', 'sys_executable': '/Users/miguel/anaconda/bin/python', 'sys_platform': 'darwin', 'commit_hash': '%h', 'traits_version': '4.6.0', 'nipype_version': '1.0.0', 'pkg_path': '/Users/miguel/anaconda/lib/python3.5/site-packages/nipype', 'commit_source': 'archive substitution', 'nibabel_version': '2.2.1', 'numpy_version': '1.14.0', 'sys_version': '3.5.4 | packaged by conda-forge | (default, Dec 18 2017, 06:40:48) \n[GCC 4.2.1 Compatible Apple LLVM 6.1.0 (clang-602.0.53)]', 'networkx_version': '2.1'}
1.0.0

[djarecka]

@mmromero thanks for letting us know.

@oesteban - I just want to double check with you, since I checked that you were changing the order in your commit. Was it just a mistake?

[oesteban]

@mmromero aren't they written in the right order here?

[mmromero]

@oesteban, I had different results when using dti_recon from the termnial and nipype DTIRecon with exactly the same configuration. So, after diving into the details I found that the gradient_matrix.txt files were different.

I modified my local DTIRecon to read the bvecs file as follow, and the results from dti_recon and DTIRecon were the same.

    def _create_gradient_matrix(self, bvecs_file, bvals_file):
        _gradient_matrix_file = 'gradient_matrix.txt'
        with open(bvals_file) as fbvals:
            bvals = [val for val in re.split('\s+', fbvals.readline().strip())]
        with open(bvecs_file) as fbvecs:
            bvecs_x = [
                val for val in re.split('\s+',
                                        fbvecs.readline().strip())
            ]
            bvecs_y = [
                val for val in re.split('\s+',
                                        fbvecs.readline().strip())
            ]
            bvecs_z = [
                val for val in re.split('\s+',
                                        fbvecs.readline().strip())
            ] 
...

[djarecka]

@oesteban the part where you read bvecs_file has different order than expected. I was checking that odf have the "traditional" x,y,z order.

I can reorder the reading part, but since I saw you changed the order from x,y,z to z,x,y some time ago, I wanted just check if you had anything specific in mind.

[oesteban]

I can't recall why is that. Probably a bug :(

[djarecka]

@oesteban ok, sorry for spamming, just didn't want to reverse your changes without checking.

@mmromero - thank you again for reporting, would you like to create a pull request with your changes?

[mmromero]

Sure, I will.

[effigies]

Hi @mmromero we're hoping to release on Monday. If you have a few minutes to submit this fix, it would be good to get it in quickly.

[mmromero]

@effigies, I don't have rights to push a branch into the repository. So I cannot create the pull request. Can you grant me access?

[effigies]

You need to fork a copy of the repository, and push to that.

See the contributing guide.

[mmromero]

I see. Since it was a very simple modification I used the web interface to create the PR: #2460