Registration of DTI to T1 in the tractograph analysis

1

Hello

I want to ask you a question. I want to use mrtrix3 to make a case report of the tractograph of the whole brain. In your tutorial, T1 was segmented and registered into DTI. Can I turn over and register DTI into T1 for tractograph analysis of the whole brain? Because if I do this, I can directly draw the ROI in T1. And I did it once according to your tutorial. I found that the TCK of the whole brain I analyzed does not coincide in the T1 background.

截屏2022-09-02 01.53.07
截屏2022-09-02 01.53.071920×1200 141 KB

If I can do what I say, what command should I use?

I hope you can help and thanks for your time,
Chen

2

Hello

I changed the steps of analysis according to my own ideas.

  1. mrconvert t1.nii *T1.mif
  2. 5ttgen fsl *T1.mif 5tt_nocoreg.mif
  3. dwiextract *den_unr_preproc_unbiased.mif - -bzero | mrmath - mean *mean_b0_preprocessed.mif -axis 3
  4. mrconvert *mean_b0_preprocessed.mif *mean_b0_preprocessed.nii.gz
  5. mrconvert 5tt_nocoreg.mif 5tt_nocoreg.nii.gz
  6. fslroi 5tt_nocoreg.nii.gz 5tt_vol0.nii.gz 0 1
  7. flirt -in 5tt_vol0.nii.gz -ref *mean_b0_preprocessed.nii.gz -interp nearestneighbour -dof 6 -omat struct2diff_fsl.mat
  8. transformconvert struct2diff_fsl.mat 5tt_nocoreg.nii.gz *mean_b0_preprocessed.nii.gz flirt_import struct2diff_mrtrix.txt
  9. mrtransform *den_unr_preproc_unbiased.mif -linear struct2diff_mrtrix.txt -inverse *den_unr_preproc_unbiased_nocoreg.mif
  10. dwi2mask *den_unr_preproc_unbiased_nocoreg.mif DTI_mask_den_unr_preproc_unb_nocoreg.nii.gz
  11. dwi2response dhollander *den_unr_preproc_unbiased_nocoreg.mif wm_nocoreg.txt gm_nocoreg.txt csf_nocoreg.txt
  12. dwi2fod msmt_csd *den_unr_preproc_unbiased_nocoreg.mif wm_nocoreg.txt wmfod_nocoreg.mif csf.txt csffod_nocoreg.mif -mask *mask_den_unr_preproc_unb_nocoreg.nii.gz
  13. mrconvert -coord 3 0 wmfod_nocoreg.mif - | mrcat csffod_nocoreg.mif - vf_nocoreg.mif
  14. mtnormalise wmfod_nocoreg.mif wmfod_norm_nocoreg.mif csffod_nocoreg.mif csffod_norm_nocoreg.mif -mask *mask_den_unr_preproc_unb_nocoreg.nii.gz
  15. 5tt2gmwmi 5tt_nocoreg.mif gmwmSeed_nocoreg.mif
  16. tckgen -act 5tt_nocoreg.mif -backtrack -seed_gmwmi gmwmSeed_nocoreg.mif -select 10000000 wmfod_norm.mif tracks_10mio_nocoreg.tck
  17. tckedit tracks_10mio_nocoreg.tck -number 200k smallerTracks_200k_nocoreg.tck
  18. tcksift -act 5tt_nocoreg.mif -term_number 1000000 tracks_10mio_nocoreg.tck wmfod_norm_nocoreg.mif sift_1mio_nocoreg.tck
  19. tckedit sift_1mio_nocoreg.tck -number 200k smallerTracks_sift_200k_nocoreg.tck
    And the following results were obtained:
    截屏2022-09-02 16.50.57
    截屏2022-09-02 16.50.571920×1200 155 KB

    I wonder if this method can work.
    I look forward to your reply!

CHEN

3

Hi 宏坤
Actually I didn’t often register T1 into DTI, which certainly will decrease the resolution of T1 image. Could you tell me which part of the tutorial you read that register T1 into DTI?

4

HI明

In fact, I first register the DTI on the tape T1, and then do dwi2response processing on the DTI
The specific operations are:

  1. flirt -in 5tt_vol0.nii.gz -ref *mean_b0_preprocessed.nii.gz -interp nearestneighbour -dof 6 -omat struct2diff_fsl.mat
  2. transformconvert struct2diff_fsl.mat 5tt_nocoreg.nii.gz *mean_b0_preprocessed.nii.gz flirt_import struct2diff_mrtrix.txt
  3. mrtransform *den_unr_preproc_unbiased.mif -linear struct2diff_mrtrix.txt -inverse *den_unr_preproc_unbiased_nocoreg.mif
    Step 1: first calculate the mat file of T1 registration to DTI
    Step 2: convert the mat file into a file used by mrtrix
    The last step is to register DTI into T1.
    If there is any error, please correct it

CHEN

5

Hi 宏坤
From the step one I noticed that you used b0 image as the follow of option -ref,
which meant you computed the transmatrix that could register T1 into DTI space.
Therefor in the step three your input should be a T1 image and which surely in contradiction with your target.
If you want to register your b0 image into T1 space, you probably should exchange the follows of the option -in and -ref, then get the transmatrix that could be used to register other diffusion images into T1 space such as b1000 in step three
Mia

6

Hi 宏坤
I noticed that maybe you could speak Chinese.
If you didn’t figure out, you could speak Chinese with me, and I will change my explanation into Chinese, hope that I could adress your problem.

7

Hi明

Thank for your question.
The -inverse option in the mrtransform command in step 3 can be applied to perform an inverse transformation according to the transmatrix of registration from T1 into DTI space.

CHEN

8

Hi 宏坤
Sorry that I miss the option -inverse, thank you for reminding me! :face_holding_back_tears:
Now I think your pipline may correct but I’m not for sure.
Maybe you could get the answer by reading some other tutorials and checking your image

9

Hi明

Thank you for your guidance！ :blush:

CHEN