Hi,
> Sorry, to be clear, I'm trying to register DWI data with a T1/T2. I only have fieldMap data to use in this process. I have a nice fieldMap to T1 string, I can get a shiftmap out of fugue for the b0, I just can't figure out how to string the shiftmap and fieldmap_to_T1.mat file I have.
I would generally recommend using BBR (which is used inside epi_reg) and avoiding fugue if possible, as the results with BBR are usually much better.
> I tried using epi_reg, with the b0 in place of EPI, T2, in place of the typical T1, but the results weren't very good.
Ah, if you don’t have a T1 then it is more complicated.
The epi_reg script will probably be failing to get a white matter segmentation, so this is something that you need to do yourself, prior to running the script. Once you’ve run FAST and generated a binary white matter segmentation image (you can binarise the pve output using fslmaths) then you can feed this directly into epi_reg using the —wmseg option. I think everything else inside epi_reg should still work, but if the registrations to the T2 image internally prove to be a problem then it will be possible to circumvent this and use flirt directly to get a BBR registration.
> The distortion correction from fugue looked much better than the output of epi_reg (even with a good WM segmentation on the T2), the problem is that I can't figure out how to use the shiftmap like I would a .warp or .mat file. Is there a file I can spit out from fugue that I can sum/string together with a flirt or fnirt output so that I can have a b0 to T1 (or MNI)?
It would be possible to do this, but I think you should try epi_reg, as outlined above, once more in order to see if you can get the best registration quality.
All the best,
Mark
> On 11/27/2014 09:36 AM, Mark Jenkinson wrote:
>> Hi,
>>
>> I would advise you to use epi_reg to achieve this.
>> Start by getting a white matter segmentation from your T2, and then use this along with your EPI images and fieldmaps. There is no need to do any registration before using epi_reg, as it will do all the necessary registrations internally.
>>
>> All the best,
>> Mark
>>
>>
>>> On 25 Nov 2014, at 17:57, Joel Bruss <[log in to unmask]> wrote:
>>>
>>> Jesper-
>>>
>>> Thank you for your response. I played around with FUGUE and was able to save out my corrected b0 and a bo_correction shift file.
>>>
>>> I have a series of registrations (T1_to_T2, fieldmap_to_T2), to which I can invert, sum and save as fieldmap_to_T1.mat (affine). How do I go about summing the bo_correction shiftmap and fieldmap_to_T1.mat? I don't see a way with convert_xfm to take a shiftmap and no matter what I do with convertwarp, anything I warp via that output (b0_to_T1_warp.nii.gz) is badly misplaced from the T1.
>>>
>>> -Joel
>>>
>>> On 11/25/2014 07:24 AM, Jesper Andersson wrote:
>>>> Dear Joel,
>>>>
>>>>
>>>>> I have DWI data collected but only one b0 image, with the same phase encoding direction. It's my understanding that I need two b0's with opposite phase encoding in order to use and apply topup. I do have a phase and magnitude fieldmap collected with the same orientation, direction and dimensions. Is there a way to apply this fieldmap to correct for distortion to the b0 and DWI data?
>>>> yes, but then you need to use different tools than applytopup. You can for example use fugue, or you can feed a 1D displacement field (shiftmap) into convertwarp if you want to combine it with registration into standard space.
>>>>
>>>>> Along these lines, could I use this fieldmap with BBR to help in alignment to a 3D T2 that was also collected?
>>>> I don’t think there should be any problems to use BBR with a T2 image. As long as your segmentation (of the T2) works well you should be fine.
>>>>
>>>> Jesper
>>>>
>>>>
>>>>> I can create the prepped fieldmap with fsl_prepare_fieldmap but I don't know what to use to apply this to the b0. epi_reg and the tools in that script look to be optimized for epi to T1 registration only. Is there a way to insert a prepped fieldmap into apply topup or use the current scans that I have to correct for distortion? Thanks for any and all responses.
|