FMRIPREP susceptibility distortion correction-Question about field map voxel and image dimensions

Dear Experts

I am reanalyzing some pre-existing datasets with resting and task fMRI data and field maps. I used FMRIPREP for preprocessing and have several questions regarding the field maps:

  1. If I have EPI 4D volumes of different tasks scanned in the same scanning session (all the different task are acquired with the same scanning parameters) but only acquired one field map, could this one field map be applied to all the different task EPI datas? Or should each EPI 4D volume have its own corresponding field map?
  2. What if their are slight differences in the scanning parameters? For instance, I have a resting state scan with TR=2.1 and a task scan with TR=2, and all the other parameters are the same, could the same field map be applied to both the resting-state scan and the task scan?
  3. What about differences in image dimension? In another existing dataset, the voxels size in all the task 4D volumes and the field map are 3.516x3.516x3mm, but the image dimension of the EPI volumes are 64x64x36, and the field map’s image dimension is 64x64x38. Could this field map be used on these EPI data?
  4. Finally, what if the voxel dimensions are different? In another dataset, the image dimension of both the EPI and field maps are 64x64x36. However, the field map’s voxel dimension was 3.516x3.516x3mm, but the EPI volume has a higher voxel resolution with 3x3x3mm. Therefore the FOV are different between the EPI and the field map, but the transformation matrix stored in the nifti make the contour of the brain overlapped pretty well among the two. Could this field map still be used for this EPI?

I have tried all of the above with the FMRIPREP pipeline, and all of them performed the susceptibility distortion correction without raising error in the report.
I would like to ask whether the above preprocessing with SDC is acceptable?
Is the field map being estimated able to undergo interpolation to deal with the dimension difference?
Or should I just give up using the field map to do SDC when the TR, voxel dimension, or image dimension are not all exactly the same?

Thank you very much!


Did you have a look at the HTML reports? There should be before/after plots for susceptibility distortion correction that should allow you to evaluate if everything went well.

Hi Chris

Yes, I did look at the HTML reports and all the above conditions I mentioned raised no error at the end of the report and all the before/after plots of SDC correction of each tasks data was reasonable to me by visual inspection.
I am just wondering whether there is a consensus on only using the field map with exactly the same voxel or image dimensions?
Or if field map could still be interpolated and therefore could deal with these dimension differences?

Thank you very much!

I don’t see why the fieldmap could not be interpolated (in fact some papers suggest using a fieldmap template in absence of real fieldmaps). If the results look good there is no reason to skip using fieldmaps.

Thank you for your response!

That was what I thought initially and hence I ran my existing datasets with the field maps with the FMRIPREP.
It was just that I came across a manual of a field map related toolbox which stated:
" For the geometric distortion correction to work, it is extremely important that the number of slices, voxel size and slice orientation of the EPI and field map are the same. Please make sure that this is examined at the scanner before data collection."
And I started to wonder whether the field map could not be interpolated to deal with the dimension difference. Perhaps the statement above has to do with the specific processing steps of that particular toolbox with the field map and distortion correction.