Inhomogeneity field estimation without reverse phase

preprocessing
#1

Dear MRtrix experts,

I’m hoping you can provide some advice on how to best account for image distortions when reverse phase B0 images are not available.

In my data (single shell, 30 directions, b=700, 6 B0 images, P->A aquistion) I see consistent distortion in the frontal lobe as well as hyper intensity in the temporal lobes. Hopefully this is obvious in this fslview screenshot:

46%20PM
Screen Shot 2018-05-15 at 12.34.46 PM.jpg1126×856 114 KB

If I understand correctly this is the sort of thing that fsl topup lives for. However, I have no reverse phase encoded B0 images.

We do collect a pair of phase map images, but am not sure if these are helpful:

14%20AM
Screen Shot 2018-05-16 at 10.03.14 AM.png1906×938 57.5 KB

51%20AM
Screen Shot 2018-05-16 at 10.04.51 AM.png1916×845 36.7 KB

Additionally, I have an MPRAGE T1 image and a rather low resolution T2 image.

Any advice on how to best deal with these distortions given the limitations of my data would be very appreciated!

Thank you for all your work,
Derek

#2

There are unfortunately no facilities within MRtrix3 for distortion correction based on field maps. But you should be able to use FSL fugue to do this.

Otherwise, I think there are ways of doing distortion correction based on the T1, but again, not within MRtrix3. I’m not familiar with the packages required for this, or how well they work, but hopefully others will be able to fill in the gaps…

#3

Thank you for pointing me in the right direction.

This thread also seems helpful: Distortion correction using T1

#4

There’s a good quantitative evaluation of the performance of multi-echo gradient echo field-mapping v.s. reversed phase-encoding EPI acquisitions for susceptibility distortion correction here.

I would be curious myself to see a demonstrated example of fugue in use, since theoretically the functionality could be wrapped by dwipreproc (if someone had the time :roll_eyes:).

#5

Theoretically, it is possible to use eddy with -field option. I tried it, but there are several pitfalls I encountered:

  1. There is a problem with robustness/precision of registration of fieldmap to DWI. The magnitude image typically does not offer good contrast, resolution and has to be precisely skull-stripped (another problem).
  2. The field map has to be invertible (necessary, but not sufficient condition is that they are sufficiently smooth and without intensity discontinuities). If it is not the case, the eddy totally fails.
    See relevant thread on FSL mailing list:
    https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1705&L=FSL&P=R22292&1=FSL&9=A&J=on&d=No+Match%3BMatch%3BMatches&z=4