Case 4 fieldmaps in GE: fmap vs dwi


I’ve run into what I believe is a contradiction on the Case 4: Multiple phase encoded directions (“pepolar”) of the latest BIDS specifications.

In GE one can not only acquire b-value = 0 diffusion weighted images (1 volume - 1 b-value=0, as you can do in Siemens), but you always have to acquire DTI volumes (7 volumes - 1 b-value = 0 and 6 b-value = 1000 (or any other b-value)). So if one wants to acquire diffusion weighted images with different phase encoding direction he needs to acquire at the very least the following two series:

Series 1: DTI volume with pepolar 0.
Series 2: DTI volume with pepolar 1.

Of these, one will be AP and the other PA, satisfying the requirements to apply this fieldmap correction case 4.

As you may see, the normal consideration of these images is DTI, see that the two of them have b values=[0, 1000, 1000, 1000, 1000, 1000, 1000]. Thus, I believe their name/folder structure should be as follows: dwi/sub-01_dir-AP_dwi.nii.gz and dwi/sub-01_dir-PA_dwi.nii.gz and they should be accompanied by their corresponding bval/bvec files.

However, according to the standard, at the very least one of them should be in the fmap directory, and have the extension _epi.nii.gz .

  • Are my assumptions correct until here? What should be the correct behavior?

Next the second problem comes with the usage of IntendedFor, B0FieldIdentifier and B0FieldSource with their json. Regardless of fmap vs dwi location of either of the files above. My goal is to use both images for fieldmap estimation, and also to correct the two of them.

As for my understanding, the two of them should have the same B0FieldIdentifier and B0FieldSource but if we want to be compatible with IntendedFor. The two of them should also have a IntendedFor list containing the two of them (on recursive pointing to the same file, and the other to the other file).

  • Is this correct?

Thank you very much for your help.

Hi Oscar, which tools/software will you use to analyse your diffusion DATA?

If you use QSIPREP, you don’t have to bother about the fmap folder, just put your two sets of DWI images in the dwi folder and the software will automatically extract the b0 images in each set to correct for SDC in the two sets.

I’m trying to do it tool agnostic… that’s why I have so many questions about the specifications lately.

In any case, I would say that you don’t really need to fill a fmap folder in that case. Most of the Diffusion preprocessing tools are able to deal with a pair a DWI sets with reverse phase encoding direction. There are still many ways to process such a dataset and any software would need configuration information to know if it should preprocess each dataset separately, concatenate of average the corrected sets, etc…

But if you wanted to use the b0 volumes to be able to correct other images, you could do the following:

  • extract the b=0 volume for each set and rename them: fmap/sub-XX_dir-{AP,PA}_epi.nii.gz
  • For each of them, add a json file fmap/sub-XX_dir-{AP,PA}_epi.json with these additional fields:
    “IntendedFor”: “path/to/images/to/be /correct/for/SDC”
    “B0FieldIdentifier”:“B0map” (and add “B0FieldSource”:“B0map” to your images to be corrected).

I will also add, for what it’s worth, that if you do not acquire fieldmaps, there is a good method of estimating the fieldmap based on synthesized data. See code here GitHub - MASILab/Synb0-DISCO: Distortion correction of diffusion weighted MRI without reverse phase-encoding scans or field-maps and paper here

Thank you very much for your input @jsein

I’ll proceed as you say then, and not to fill up the fmap folder since it is not needed in this case. Just one more question in this regard, should the IntendedFor tag be used? since it would be recurrent.
If it has to be used, I believe it should be like this::
File 1: ses-X/dwi/sub-01_dir-AP_dwi.nii.gz
File 2: ses-X/dwi/sub-01_dir-PA_dwi.nii.gz
Json 1:

   "IntendedFor": [

Json 2:

   "IntendedFor": [

See that both JSON are the same in regards to the IntendedFor tag.

I know other vendors such as Siemens do allow the acquisition of only the b0 images, so in those case they should indeed go to the fmap folder.

Finally, I am assuming the same approach can be followed for other EPI sequences such as fMRI, right @jsein ?

Thanks @Steven for you input, I’ll carefully review the paper you mention, however, I already have this data.

@Oscar : Yes, you got it right!

For this case:

You would then have for instance one b0 in AP and one full DWI set in PA.
In that case I would do:

  • ses-X/fmap/sub-Y_dir-AP_epi.nii.gz with ses-X/fmap/sub-Y_dir-AP_epi.nii.json having:
   "IntendedFor": [
  • dwi: ses-X/dwi/sub-Y_dir-PA_dwi.nii.gz, ses-X/dwi/sub-Y_dir-PA_dwi.bvec, ses-X/dwi/sub-Y_dir-PA_dwi.bval

with ses-X/dwi/sub-Y_dir-PA_dwi.json having:

1 Like