Hi Hedok
We use the TBSS-style alignment method and use the FA map to align
non-linearly to FMRIB58_FA. unfortunately we weren't able to get IRTK
working all that well on our system so i have been using SPM and getting
very good alignment of tracts.
i am now looking at whether there are major differences in the output of
FNIRT and SPM in this regard.
cheers
Scott
Hedok Lee wrote:
> Hi All,
>
> I don't want to complicate this issue, but I'm also interested in
> registering diffusion with anatomical image in the absence of field
> map to correct image distortion retrospectively. I read some where in
> the list where Jesper was saying it's "too permissive" of an approach
> to register EPI with anatomical image this way. This is because the
> correction should be done along phase encoding direction only.
>
> From what I understand, EPI image is distorted along phase encoding
> direction(typically vertical axis) making a sphere look like an egg
> shape (ellipsoid). To register the ellipsoid with the sphere, I think
> there are more than one ways to do this. One of the ways is to shrink
> the ellipsoid along the y-axis to match the sphere. Alternatively, it
> can also be shrunk by shrinking along x,y, and z axis. Thus non-linear
> registration between EPI and anatomical image should be done with some
> restrictions, and I'm not sure if FNIRT has an option to warp along
> prescribed axis. Please let me know if this is feasible with FNIRT.
>
> There are papers written about EPI distortion correction without use
> of field map like this one
>
> Geometric Distortion Correction of High-Resolution 3 T Diffusion
> Tensor Brain Images
> Siamak Ardekani1 and Usha Sinha, MRM 2005
>
> Hope this helps?
>
> Hedok
>
>
> Gwenaëlle DOUAUD wrote:
>> Hi Anastasia,
>>
>> I think Jesper would have a better insight on the question, but I
>> don't think that registering dw images to T1w images should be a
>> problem.
>>
>> Having said that, I am not surprised that non-linearly registering b0
>> images to T1 scans don't work (see previous threads on the question).
>> There is not enough contrast in this image to properly match the
>> structures. Try generating the FA map and then fnirt it to the T1
>> with the default settings. It should work fine.
>>
>> Cheers,
>> Gwen
>>
>>
>>
>>
>>
>> --- En date de : Mer 20.8.08, Matt Glasser <[log in to unmask]> a écrit :
>>
>>
>>> De: Matt Glasser <[log in to unmask]>
>>> Objet: Re: [FSL] FNIRT bzero to T1
>>> À: [log in to unmask]
>>> Date: Mercredi 20 Août 2008, 21h05
>>> Nonlinear registration like FNIRT is not really appropriate
>>> for cross
>>> modality registrations. It is better to properly unwarp
>>> your diffusion
>>> weighted images using a field map and fugue or some other
>>> method and then
>>> use rigid body registration (FLIRT -dof 6) to register each
>>> subject's FA map
>>> to their T1 scan.
>>>
>>> Peace,
>>>
>>> Matt.
>>>
>>> -----Original Message-----
>>> From: FSL - FMRIB's Software Library
>>> [mailto:[log in to unmask]] On Behalf
>>> Of Anastasia Ford
>>> Sent: Wednesday, August 20, 2008 12:39 PM
>>> To: [log in to unmask]
>>> Subject: [FSL] FNIRT bzero to T1
>>>
>>> Hello,
>>>
>>> I am wondering if anyone is using FNIRT to register DWI
>>> volumes and
>>> anatomical scans. I tried doing so by changing the configuration
>>> files provided
>>> in the new FSL
>>> package but have not found a set of parameters that would enable a good
>>> registration.
>>> Sincerely,
>>>
>>> Anastasia
>>>
>>
>>
>>
>> _____________________________________________________________________________
>> Envoyez avec Yahoo! Mail. Une boite mail plus intelligente
>> http://mail.yahoo.fr
>>
--
========================
Scott Kolbe
MS Imaging Group
Howard Florey Institute &
Centre for Neuroscience
University of Melbourne
VIC, Australia, 3010.
ph: +61 3 8344 1887
email: [log in to unmask]
website: http://www.neuroimaging.org.au/index.php?id=383
|