3D T2 fasciculus in 3D T1 reconstruction

[synic]

Hello,

I try to assemble my 3D logituninal fasciculus reconstruction in T2 with my 3D brain reconstruction in T1. How can I readjust the T1 with the T2 in order to have a valid 3D representation ?

thanks a lot for solution,

michel

[Jean-Francois Mangin]

Unfortunately,
if you did not acquired additional data (phase map) to correct for the EPI distortion relative to the high resolution T1, there is no simple solution. We plan to deliver such a correction in a near future, but without the phase map, you need to look elsewhere. For instance you could look for a non linear registration method, the best one for this purpose using the fact that the main distortion is in the encoding direction.

[Manik]

Hi Michel,

for this, you need to compute the transformation from the T2 referential to the T1 referential, with a registration program. There are some of these programs in "registration" processes in BrainVisa ("recalage" in french), but I don't know if they work with two different contrasts like T1 and T2.

[Jean-Francois Mangin]

Hi Michel,

for this, you need to compute the transformation from the T2 referential to the T1 referential, with a registration program. There are some of these programs in "registration" processes in BrainVisa ("recalage" in french), but I don't know if they work with two different contrasts like T1 and T2.

The method based on mutual information should work (AimsMIRegister), if you are lucky, but I am not sure of the best parameters (there is a post somewhere on this subject) . Unfortunately the EPI distortion will not be overcome that way.

[synic]

thanks,

I try this now !

[denghien]

Hi,

You must coregister datas between them. For this, there is 2 command lines (there is a process in BrainVISA but it seems to me it bugs, it will be ok in the next paketage) :

With command lines, you can do somethings like that :

--> AimsManualRegistration -f from.arg -t to.arg -o fromTOto.trm

* It is a manual registration where you draw a ROI graph with 3 voxels for each volume. Points must have the same meaning. For instance :
from.arg --> ROI graph for T2
to.arg --> ROI graph for T1
fromTOto.trm --> file of motion, it is a simple text file. You can use it to change referentiel into Anatomist or resample datas

--> AimsMIRegister -r T1.ima -t T2.ima -o T2TOT1.trm
This command uses Mutial Information to register.
T1.ima -->T1 volume
T2.ima --> T2 volume
T2TOT1.trm --> motion file like fromTOto.trm

Note : Volumes are not resampled. If you want resample data with information of file motion (trm), you can use this :
AimsLinearResample -i T2.ima -o rT2.ima -m T2TOT1.trm

Note commands coregister only and don't correct distorsions ! I don't known really what you want.

Tell me if you obtain a good result, please


Isa

[synic]

Hello,

it's seems to work really good when I do this :

In this case I register the "nobias T1 MRI" on the T2 MRI. After, I only have to make my 3D reconstruction on the new (registered) nobiasT1 MRI.

(Jean-Francois Mangin's old message in the forum)
AimsMIRegister was developed to register TEP and MRI exams using rigid spatial transformation and mutual information as optimization criterion.

Commandline to register mri1.ima with mri2.ima:

AimsMIRegister -r mri1 -t mri2 --seuilref 0.05 --seuiltest 0.05

Two transformations are estimated: mri1_TO_mri2 and mri2_TO_mri1
The registration can be evaluated visually using anatomist in fusion mode.

Commandline to resample mri2:

VipSplineResamp -i mri2 -t mri1 -d mri2_TO_mri1 -o rsp_mri2 -r t -w t