orientation of phase-encoding reversed DWI image

[mcraig-ibme]

I am processing some rodent data which includes forward and reversed phase-encoding direction DWI scans. I am finding that the reversed scan does not display consistent orientation relative to the other DWI and anatomical scans when viewed in an orientation-aware Nifti viewer such as fsleyes. The PE reversed image appears inverted in the Y direction which seems consistent with the PE reversal not being accounted for.

Looking at the acqp files for the two scans, the key parameter seems to be:

##$ACQ_scaling_phase=1
(for forward scan)

##$ACQ_scaling_phase=-1
(for reversed scan)

I may be able to provide a sample dataset (but need to confirm with researcher first)

I would expect the two images to view with the same orientation, possibly with a difference in the Nifti affine to account for the different internal data ordering.

This is how the forward scan displays:

And this is the reversed:

This is on Centos Linux using current Github code for brkraw - thanks for any help and can provide further details, e.g. Nifti headers as required.

[dvm-shlee]

@mcraig-ibme Thank you for reporting this issue. Regarding the reversed phase-encoding, not sure if there are any parameters in visupars already handling this. Might need to investigate the dataset.

Could you share the sample dataset once the researcher confirms sharing? also could you let me know what PV version you are using?

[mcraig-ibme]

I have sent a link to a dataset via email, hope this helps.

[jeremie-fouquet]

Hi. We also implemented reversed phase-encode EPIs for fMRI and dMRI. Bruker suggested we not only set ACQ_scaling_phase=-1, but also modify the RecoRelations.c file to take into account the reversed image orientation correctly. I have not tested it, but I guess that by only changing ACQ_scaling_phase, the data would be upside-down even in the ParaVision display. I would argue this issue should be solved outside of brkraw.

I'm not sure yet of the guidelines for sharing Bruker sequences, but you can contact Bruker or me if you want more details. Basically, we used a RecoConjQNegFilter to flip the data and had to change the sign of the offsets in the phase encode direction.

[dvm-shlee]

@mcraig-ibme @jeremie-fouquet The 'VisuSeriesComment' in visu_pars appears to be crucial for resolving this issue. The value is 'Cardiff_Standard' for forward scans and 'Cardiff_Reverse' for backward scans. Interestingly, this parameter isn’t listed in our protocol, but 'ACQ_scaling_phase' seems to be functional.

Given that each user might have their own modified sequences, I'm considering using a plugin module for conversions in brkraw, instead of the default setup. This approach would allow users more flexibility to manage unique sequence issues.

The version scheduled for release on May 15 will support plugin functionality with command-line options. I'll prepare some example cases using your images. Since the dataset isn’t public, I will share the examples directly with you via email once the new version is available.