fsl doesn't work anymore on DTI bruker2nifti stacks with more than 1 repetition

[emicotti]

Hello,
I have worked for many years using bruker2nifti to convert DTI Bruker raw data and then calculating Diffusion Tensor with fsl dtifit command.

A few weeks ago we have installed a new workstation. Trying do the usual DTI pipeline we noticed that fsl showed only half of the data. Digging into the problem we have discovered that fsl cannot "see" anymore the second repetition (we do two repetitions to correct deformation before averaging images).

The team of fsl told us that it happens because the two repetitions are stored in the nifti header as "5th dimension".
Old fsl release "flatten" any dimension above the 4th but recent realeases simply ignored all the dimensions higher than 4.
Do you think it is possible to make bruker2niffti assign the repetitions to the 4th dimension rather than 5th dimension?

Thank you very much.
Best
Edoardo

[gdevenyi]

Hi, as per #67 this project is no longer being actively developed by its original maintainer. I suggest looking at https://github.com/BrkRaw/bruker

[SebastianoF]

Thanks @gdevenyi !

@emicotti Glad you had been using bruker2nifti over the past years!

I'll try to see if I'll ever have time to look into it. In the meantime you can either use https://github.com/BrkRaw/bruker as suggested, or find someone to take over its maintenance.

[emicotti]

Thank you all for answer and suggestions.

[SebastianoF]

You have other two options:

1. downgrade fsl to the previous version.

2. add a small post-processing before feeding the converted images into fsl.

$ pip install nilabels, nibabel

import nibabel as nib

from nilabels.tools.aux_methods.utils_nib import set_new_data


def squeezer(nib_image):
    data = nib_image.get_data()
    new_data = data.squeeze()
    return set_new_data(nib_image, new_data)

    
if __name__ =="__main__":
    img = nib.load(<your path to nifti here - the one returned by bruker2nifti>)
    squeezed_img = squeezer(img)
    nib.save(squeezed_img, <your path to squeezed nifti here - the one to feed fsl>)

[SebastianoF]

By "convention" the fourth dimension of a NifTi image is left for the temporal evolution, and the fifth for the weights of the sequence or any other value that has no "temporal" component.

Even if I had time, it would be unwise to change bruker2nifti output because of how fsl, down the line, now works.

You understand other users may ingest the data into other software that use the five dimensional convention. So what solves the problem for you may create problems to them. I would recommend adding a fsl-pre-processing step as part of your pipeline.
Thanks for your contribution, as I am sure this issue will be useful to other users!