Issue with EPI correction


I have kind of an unusual problem regarding the preprocessing DWI data acquired on a 3T Siemens Scanner:

Prior to the actual DWI sequence (b-value 1000, 60 directions), one image in AP and another in PA-direction was acquired with the EPI distortion correction using FSL topup in mind.
Unfortunately these two images were acquired with the ‘PreScan Normalization’ option for Siemens MRIs enabled but during the DWI-Scans this option was disabled.
If I want to use FSL topup and eddy to correct for the EPI distortions, it gives me unsatisfactory results due to the different intensity scaling of the AP/PA-images and the DWIs.

How would you go about this issue? Has anybody struggled with this before and knows a solution?

Is there maybe a possibility to adjust the intensities of the AP/PA images to the DWI so that topup does a better job?

I would be glad about any help/advice.

Thank you very much!

Hi Tom,

topup has the --scale option to address exactly this issue. Unfortunately, at the moment there is no way to pass extra options to topup from the dwipreproc script. I will update dwipreproc script to allow this in a minute.


I have added support to pass custom topup options to the dwipreproc script:

Once these changes have been approved and merged with the master branch, you should be able to get the desired effect by rerunning dwipreproc with the option -topup_options "--scale=1" (that is, after running git pull in your mrtrix3 source directory).


@Tom_M: give it a try with the current master (first run git pull in your source directory) and then add -topup_options "--scale=1" to your dwipreproc call. Let us know if it solves your problem.

@bjeurissen: Thank you so much for the fast reply!!!

Unfortunately at our institute software updates have to be done by the IT. This might take a while…

In the meantime I try to run topup with the --scale option manually outside of mrtrix to see if it works on my data.
How many images should I feed into topup when I have only 1 AP and 1 PA-image (with the PreScan Normalisation Option) and the actual sequence (68 volumes, without the PreScan Normalisation Option)?
I thought about setting the AP/PA-images + 3 b0-images from the DWIs as --imain to re-scale the images to a common global mean.
Would this be appropriate? I am not sure if it is a problem that I would use only 1 image in PA and 4 images in AP direction!?

Thank you!

I highly recommend using an equal amount of AP and PA images. Probably the safest option in your case is to use the single PA scan + the first b=0 scan from your actual sequence (AP).

Note that there is no requirement for the images from which you calculate the displacement map and the images to which you apply these displacements to have the same intensity scaling. Topup is estimating a displacement map, which is completely independent of the intensity scaling. This displacement map can then be applied to another image regardless of the scaling