Dear Jesper,

Thanks a lot, this helps.
I will try what you suggest on my already collected data and report 
back, as I don't have b0 with opposing PE.

All the best wishes,
Barbara



On 16/10/2015 15:31, Jesper Andersson wrote:
> Hi again,
>
> I forgot to mention in my earlier reply that in the, unfortunate, case that you have already acquired your data and you don’t have data with opposed PE directions you _may_ be able to use your T2 structural.
>
> You would then pass your EPI-b0 and your T2 into topup and specify a very low readout value (implying no distortions) for the T2 volume. If the contrast of your T2 struct and your b0 is similar enough this might give you a reasonably good estimate of the field. I have seen this work in the past.
>
> That field you can then pass into eddy to the --topup parameter.
>
> For the future I strongly recommend acquiring b0 with opposing PE as part of your DTI protocol.
>
> Jesper
>
>> Dear Barbara,
>>
>> it is not sufficient to pass a corrected b0 image to eddy. You need to pass it a susceptibility field as calculated by the FSL tool topup. topup requires at least two volumes acquired with different PE-directions in order to estimate the field. This would typically be two b0 images acquired with PE A->P and PE P->A, but can also be for example two b0 acquired A->P and L->R. We also recommend using the first b0 images acquired as part of your DTI protocol as the first image of the pair that is passed to topup.
>>
>> There is more informations and recommendations on the wikihelp.
>>
>> Jesper
>> On 16 Oct 2015, at 14:44, Barbara Kreilkamp <[log in to unmask]> wrote:
>>
>>> Dear FSL list,
>>>
>>> Could you please advise me on how to incorporate a b0 geometric distortion correction procedure?
>>>
>>> My workflow has been like this:
>>> (1) I use eddy_correct (flirt with 6 dof) on the b0 images
>>> (2) I take the mean of these co-registered b0 images
>>> (3) I extract this mean b0 image and register it to the T2-w structural scan I have (using Bspline co-registration in Slicer)
>>> (4) I bring the image to diffusion space (running flirt -dof 6 -nosearch on the new b0 to bring it to the original raw first b0 space) and merge with the raw DWI, now I run eddy.
>>>
>>> I noticed in some participants the b0 and DWI images don't overlap very well after running eddy, especially where the geometric distortions had been previously. Although for all DWI the eddy currents and gross motion is corrected for beautifully in eddy.
>>>
>>> Could you please explain why eddy does not seem to perform very well in terms of non-linear alignment of DWI to my new b0 image?
>>> Would you recommend I run my steps in a different order?
>>> Would you please point me to a publication on eddy?
>>>
>>> Thank you so much,
>>> Best wishes,
>>> Barbara