Image registration for diffusion MRI

1

Hi, Experts,

My aim is to compare the fiber and FA values among different individual. For the fibers of each individual are quite different, I wanna to register the fibers to MNI152 space. At the same time, I wanna to preserve the FA values of each fiber.

As far as I can think, there’re two ways to convert brain fiber tracts to template space: 1) Generate Fiber tracts and the corresponding diffusion attributes (such as FA, MD…) of each fiber, then warp the fibers to the template space. 2) register T1 and FOD data (or DWI data in general) to template space, then generate the fiber tracts and the corresponding diffusion attributes.

1 . About warp tck file to template space.

The previous topic is here. I find it that the warp image of data1 does not cover the whole brain tracts. It is the same with “synantsInverseWarp.nii” which is created by ANTS.

ANTS 3 -m CC[MNI152_T1_1mm.nii, t1_mpr_ns_sag_iso.nii.gz, 1, 5] -t SyN[0.5] -r Gauss[2,0] -o t1_disease_liuyan_to_MNI_synants.nii -i 30x90x20 --use-Histogram-Matching

As shown in Fig 8.1-8.4, compared with data1, the whole brain tracts of data3 are covered by warp image. So, I should say the problem list on the first command ANTS. Does it mean the parameters of the ANTS is not suitable for the data? Is it the same to warp FA image as warp the fiber tracts?

2 . About register diffusion MRI to template space.
I learned that spatial normalization of diffusion data is not recommended, and never transform FOD data (or DWI data in general) .

However, I find an Image that shows the registration for diffusion MRI, (see Fig 9): Designed explicitly for inter-subject alignment of fibre orientation distributions as provided by CSD and variants.

It seems that I should be able to achieve my goal through the above two methods. So, I am confused now.

Many Thanks,
Chaoqing

Fig 8.1 ( DTI_2013_act_0.5M.tck && T1_to_MNI_synantsInverseWarp ) – Data1

5M_&&_warpMNI152
5M_&&_warpMNI1521855×1056 680 KB

Fig 8.2 ( DTI_2013_act_0.5M.tck && MNI152_T1_1mm.nii.gz ) – Data1

nii
nii1855×1056 779 KB

Fig 8.3 ( DTI_30_average-6_act_sift_0.2M.tck && t1_disease_liuyan_to_MNI_synantsInverseWarp.ni ) – Data3

2M_&&_warpMNI152
2M_&&_warpMNI1521855×1056 658 KB

Fig 8.4 ( DTI_30_average-6_act_sift_0.2M.tck && MNI152_T1_1mm.nii.gz) – Data3

nii
nii1855×1056 694 KB

Fig 9.( Comes from MRtrix3 homepage: “Image registration for diffusion MRI” )

registration
registration750×410 78.9 KB

2

Dear @SuperClear,

I am tempted to ask a principial question: is here a reason to stick with quite out-of-date tensor metrics such as FA? Mrtrix3 offers much up-to-date and well documented fixel-based analysis pipeline.

On the other hand, if you for some reason want to stick with older tensor-based metrics, why not to use quite well established and still in community appreciated method TBSS?

Antonin

3

Hi, Antonin,

Thank you for your advice!

I really ignored fixel-based analysis pipeline in Mrtrix3 documents. I should detailly read Mrtrix3 documents.

I wanna to generate Spatial Statistics of the brain fiber tracts in MNI152 space. I should try to use TBSS method.

I will test TBSS and fixel-based analysis pipeline of Mrtrix3.

Thanks again,
Chaoqing