[Mristudio-users] motion correction makes sense?

[susumu mori]

HI Durai,

Yes, if the brain rotates, the contrast changes and therefore simple
rotation registration is, strictly speaking, not accurate.
There are papers discussing this issue and we usually solve this by
rotating the gradient vector together.

For example, we apply [1, 0, 0] gradient in the first image. In the second
image, we apply the same direction but the brain was rotated. In the
post-processing, we re-rotate the brain to register the second brain to the
first brain. In this case, the gradient orientation of the second
"re-rotate (registered)" image is no longer [1, 0, 0]. So we apply the
rotation not only the image but also the gradient orientation.

In reality, typical DWI data should have rotation less than 1 degree within
typical 5-10 min scans. So I expect this won't be a big issue, but if the
brain does rotate more than 1 degree, the gradient re-orientation does make
difference. However, if the subject moves 2-3 degree, these data are
definitely outliers, which often suffer from slice-by-slice registration
errors WITHIN one 3D volume (you can't correct it unless the rotation is
purely within the 2D imaging plane) and signal dropout. So, the issue is
usually not as simple as applying the gradient re-orientation and become a
happy person.

We are working on a comprehensive QC report to deal with this issues. I
think it will come out this year but you can have a working version in the
latest DtiStudio. Hangyi, is it correct? Do we have Yue's QC reporting
function already in DtiStudio?


On Mon, Jun 24, 2013 at 1:44 AM, Durai Arasan <durai23 at gmail.com> wrote:

> Hi all,
>
> A general question about motion correction in diffusion imaging. In fmri
> this is a commonly practiced preprocessing step where the new signal, say,
> F' is obtained by a spatial shift of F, signal acquired when movement
> occurred:
>
> F' = F*m
>
> where m = m(x,y,z), the motion matrix.
>
> But in diffusion this is clearly not the case as the signal in each
> direction is a function not only of the position of the brain but also of
> individual brain geometry. Therefore, new signal D' will be a *transform*(not shift) of signal D, acquired when movement occurred:
>
> D' = D*t
>
> where t = t(x,y,z,brain_geometry), the transform factor.
>
> So unless there is an estimate of t, I don't think any motion correction
> is going to be useful in diffusion, certainly not a simple rigid body
> adjustment. Am I wrong?
>
> Thanks
> Durai
>
>
>
> _______________________________________________
> mristudio-users mailing list
> mristudio-users at mristudio.org
> http://lists.mristudio.org/mailman/listinfo/
> Unsubscribe, send a blank email to:
> mristudio-users-unsubscribe at mristudio.org
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mristudio.org/pipermail/mristudio-users/attachments/20130624/f5dc9b57/attachment.html