DTI registration

[synic]

Hello,

I have a pb to register DTI.
I made a template in matlab, in order to have a normalized DTI. After I register my original DTI on this normalized DTI (In order to reorient my DTI without loss of information), but I loss direction information...
Do you know a method for this ?

thanks a lot


michel

[Jean-Francois Mangin]

I have some problem understanding what you try to do. The spatial normalization of DTI is a very complex problem. Either you use the T2-weighted image, or maybe the FA image, otherwise, if you want to use the full tensor, you need new similarity measures. A few groups have addressed this point in literature (see maybe D. Jones or C. Davatzikos groups)....

But what is the link with brainVISA?
As far as I am aware, nothing like that is provided for the moment?

[synic]

Thanks for your response,

Excuse me, I was not clear...
I just want to create a typical fasciculus... (in order to insert it in my normalized T1 3D reconstruction).
I thought that AimsMIRegister only reorient images but it modify too all the informations of the DTI (that's exactly the pb).
I just want to simply register my DTI (with brainvisa) with a normalised DTI.

[Yann Cointepas]

I may have missed a feature of AimsMIRegister (it is sometime hard to follow all what is going on) but I think that only the image is reoriented. It may loose the orientation information during this process (several old commands do not copy the header of the source image to the destination image). You can get this information back by manualy editing the source image *.minf file and copying the appropriate information (see http://brainvisa.info/forum/viewtopic.php?t=58 for more details). But in order to process your reoriented data, you need, at least, to find a way to apply similar deformation to the gradient orientations.

[Jean-Francois Mangin]

I still do not understand what you try to do. What do you call DTI, the raw data?
In that case you are right, you can not normalize them without changing the associate gradient direction. But I think you should think another way. Just normalize the T2-weighted images, and apply the normalisation to the fascicle image. Finaly average all the fascicles in Talairach space.

BUT, do not use AImsMIRegistration for that, I am afraid it is dedicated to rigid body motion, not to affine registration. In general, do not try to find normalize tools inside brainVISA for the moment, it does not exist. Use SPM or MRItotal.

[synic]

Thanks a lot for theses informations

michel