Dear Mark,
> I am trying to correct EPI distortions in a DWI dataset. Unfortunately I have no B0 field map or reversed polarity blip acquisitions to estimate the distortion field, so I cannot use fugue or topup. I am therefore forced to use image registration methods.
next time ...
>
> I am currently trying to use fnirt to register the B0 (non diffusion weighted) image to an undistorted T2-weighted image with the same contrast. Unfortunately, the EPI distortions cause signal pileup in some regions, which messes up the contrast matching (in particular around the anterior temporal lobes). The resultant warped image is pretty messy in these regions. I explored some paramaters of fnirt. I have tried using the --intmod=local_linear flag with a high-res bias field. I have also tried using a higher resolution for the warp. Unfortunatly nothing so far gives me satisfactory results.
>
> I am considering two new approaches, which I am hoping someone here can comment on:
>
> 1. Is there a way to constrain fnirt so that it only warps in one direction (i.e. the phase-encoding direction)? In the warped B0 images, the regions with high pileup veer off laterally. If I can constrain the algorithm so that non-phase-encoding directions are not warped, I can prevent this happening and hopefully improve the registration.
>
> 2. Is there a way to flatten the image, so that intensities in the pileup regions are closer to that in the reference (undistorted) T2w image? I have tried doing tissue segmentation and only taking values that correspond to the gray and white matter, excluding CSF, (where most of the signal hyper-intensities occur). This does help the EPI distortions a small amount, but it also introduces new warping artefacts throughout the rest of the image (presumably because of the loss of other structural details critical to calculating the warp for the rest of the brain).
topup is designed to allow you to use any pair of images as long as the translation field->distortions is different for the two images. I have for example successfully corrected images with the same blip-direction but where the bandwidth was a little lower in one than in the other.
Hence I think, I haven't tried it though, you should be able to get reasonable results by using your b=0 EPI and your T2 image in topup, but where you set the readout-time for the T2 to some very small value (I'm not sure if it will accept zero or not). It shouldn't really matter what you set your PE-direction to be in the T2 image.
I _think_ this should work. Just how well it will work will depend on how well the intensities in the T2 matches the b=0 EPI.
Give it a whirl and let me know how it works.
Jesper
>
> Any assistance that can be offered would be gratefully achieved. Or if anyone can point me to references to alternative methods that deal with this problem, that too will be very helpful.
>
> Many thanks
>
> Mark
>
> --
>
>
> Dr. Mark Drakesmith
> Research Associate
>
> Caridff University Brain Research Imaging Centre (CUBRIC)
> School of Psychology
> Cardiff University
> Park Place
> Cardiff
> CF10 3AT
>
> Tel: +44 (0) 29 2087 0354
> Fax: +44 (0) 29 2087 0339
>
|