Dear Julia,

I would use the same style of registration as described in the registration practical
except that for fnirt you will need to use the --ref option (point it at the JHU_MNI_SS_T2)
and stick with the T1_2_MNI152_2mm.cnf config file - since this should work if your
reference and input images are of the same modality.

Then use invwarp to calculate the inverse warp and finally use applywarp to
transform the JHU_MNI_SS.d images back into your native space.  Once you've
done this you can look up values in FSLView or with fslmeants.

All the best,
	Mark


On 20 Apr 2012, at 14:57, Julia Buhlmann wrote:

> Hello,
> 
> I just started working with fsl and I have a question regarding co-registration.
> 
> I have a T2 MRI of a patient (pat_t2) and I would like to get the tensor matrix at certain points. As I do not have a DWI from the same patient I would like to use the JHU-MNI-SS DTI atlas.
> 
> From this atlas I have the tensor image (JHU_MNI_SS.d) which contains 6 image volumes of the tensor matrix, a co-registered T2 MRI (JHU_MNI_SS_T2) and a co-registered T1 MRI  (JHU_MNI_SS_T1).
> 
> I have tried to co-register pat_T2 as input to JHU-MNI-SS_T2 as reference, so that pat_T2 is registered with the JHU_MNI_SS.d. I used the following commands.
> 
> - bet pat_T2 pat_T2_brain -f 0.35
> - bet JHU_MNI_SS_T2  JHU_MNI_SS_T2_brain -f 0.35
> - flirt -in pat_T2_brain -ref JHU_MNI_SS_T2_brain -dof 12 -cost corratio -out flirted_pat_T2 -omat pat_T2_to_JHU_T2.mat
> 
> After this I wanted to use fnirt as well but I am not sure if this is the correct way and about the additional parameters to use.
> The flirt registration needs to be improved.
> 
> 1) Could you describe me how I can find the best registration? So far I followed the inter-subject registration at “registration practical” on the fsl web site (http://www.fmrib.ox.ac.uk/fslcourse/lectures/practicals/reg/index.html).
> 2) How can I find the certain points (to read the tensor) in the pat_t2 image after my co-registration, especially if I have to do a non-linear registration?
> 3) Is there another/better way to get the tensor values out of the DTI-atlas (JHU_MNI_SS.d) at certain points in the pat_t2 image?
> 
> Thank you for your help.
> 
> Julia
>