Dear FSLers,
is there any news on this topic?
We are also wondering whether we could adjust bvecs to motion corrected data. Especially since we
also have patients` DTIs, which are most likely to be distorted by motion.
Any experience on the differences of adjusted/ non-adjusted data yet?
Regards,
Mareike
--
Mareike Menz
NeuroImage Nord Lübeck
University Medical Center Hamburg-Eppendorf, Bldg. S10
Martinistr. 52
20246 Hamburg
Germany
+49 40 42803 3683
[log in to unmask]
--
Tim Behrens wrote:
> Hi there
> On 9 Dec 2006, at 01:03, Bennett Landman wrote:
>
>> Good afternoon,
>>
>> 1) We wrapped FSL FLIRT with a Matlab GUI to perform motion correction,
>> tensor adjustment*, and tensor estimation. Of course, you can use only
>> the
>> first two of these steps if you only want to do motion correction. The
>> tool
>> is called CATNAP and available from:
>> http://iacl.ece.jhu.edu/~bennett/catnap/catnap.shtml Although it
>> primarily
>> handles complexities with diffusion weighting coordinate system on
>> Philips
>> scanners, you can use it with Analyze data from any other system provided
>> that you can specify the diffusion weighting directions relative to the
>> slice. These are simple steps, but the wrapper makes it easier for our
>> users
>> to manage the multi-volume data. (Comments welcome.)
>>
>
> I would be interested to know how much of a difference this step makes
> (i.e. the rotation of the bvecs with the image). I must admit that we
> have not done it previously, as we thought that this effect would be
> very small in comparison to other effects of motion that we cannot
> correct for (such as dropout effects, or the interaction between motion
> and field inhomogeneity effects). However, we don't have evidence for
> this. Do you find that it makes a big difference? Can you show images
> with and without the bvec correction. If it looks like a good idea, we
> would be happy to write a little tool to rotate the b-vectors after
> running eddy_correct - as martin has said, eddy_correct is also just a
> wrapper for flirt and we already store the rotation matrices.
>
>
>> 2) I doubt that you need to discard DTI data due to affine motion. If you
>> are not using an EPI distortion correction tool, it is possible to get
>> fairly large translations that have nothing to do with motion. For a
>> first
>> pass, I would inspect the images and reject motion artifact (e.g., in
>> DTIStudio). More robust approaches detect and remove the influence of
>> outliers (e.g., RESTORE).
>>
>
> I agree with this - it would be dangerous to reject images just on the
> amount of absolute detected motion.
> Some motion effects are easy to correct, but others are much harder
> (e.g. the ones I mention above). You should check through all of your
> data for motion-related effects ( e.g., you might see that a large
> potion of a single slice looks black).
>
> Cheers
>
> Tim
>
>
>
>
>> * When motion correcting diffusion weighting data, it is necessary to
>> update
>> the diffusion weighting directions (b-vecs) to reflect the true
>> relationship
>> between anatomy and sensitization directions. For a thorough
>> discussion of
>> this, please see Jon Farrell's site:
>> http://godzilla.kennedykrieger.org/~jfarrell/software_web.htm
>>
>> Regards,
>> Bennett
>>
>> -------------------------------------------
>> Bennett Landman, [log in to unmask]
>> Ph.D. Candidate, Biomedical Engineering
>> Johns Hopkins University School of Medicine
>>
>>
>> -----Original Message-----
>> From: Chunchun Ni [mailto:[log in to unmask]]
>> Sent: Friday, December 08, 2006 5:29 PM
>> Subject: Motion correction in DTI data
>>
>> Hi, :)
>> I have questions about motion correction in DTI data:
>> 1) How to detect/correct motion? Can Eddy current correct motion? Can SPM
>> realignment detect and correct motion? Some other tools suggested?
>> 2) What's the criteria to discard images that have too large motion? In
>> fMRI, some people use 2mm or 2 degree. How about DTI?
>
|