Skip to first unread message
Clara
May 8, 2017, 2:10:30 PM5/8/17
to ExploreDTI
Hi ExploreDTI Group,
I am using ExploreDTI for the eddy current and EPI correction on my diffusion data. I use a T1 image to register my DWI data to the anatomy in this step as well. I have previously outlined my steps, and it works great for all my data except for 1 scan from 1 participant. I am hoping someone might spot what is going wrong with this participant.
To supply full context, here are the steps I followed:
1) Converted .nii.gz files (DWI and T1) to .nii using freesurfers mri_convert
2) Converted *.bval/*.bvec to B-matrix.txt file.
3) Used Flip/Permute dimensions of 3D/4D *.nii files with the default parameters.
4) Shuffled DWIs so that the b0 weighted volumes were in the beginning.
5) Calculated DTI *.mat file.
6) Corrected for subject motion & EC/EPI distortions
a) Registering to other data to do the EPI correction (non-rigid)
b) Defining the image type as FA
c) Setting the deformation axes as 1 0 0
d) Leaving all other settings as they are per default in ExploreDTI v4.8.6
I have 2 scans from the participant in question, because we conducted a longitudinal study with ~4 weeks between scans. Please find the 2 corrected FA images attached (exported from the trafo.mat file).
I have loaded both the T1 and the FA into FSLVIEW and what I took screenshots of is the colored T1 image and the grayscale FA map. In the scan1_brain_ok image, you see very little of the T1 under the FA. However in the scan2_brain_squashed image, you can see a ring of T1 around the whole FA map.
Any suggestions?
Thank you in advance!
I am using ExploreDTI for the eddy current and EPI correction on my diffusion data. I use a T1 image to register my DWI data to the anatomy in this step as well. I have previously outlined my steps, and it works great for all my data except for 1 scan from 1 participant. I am hoping someone might spot what is going wrong with this participant.
To supply full context, here are the steps I followed:
1) Converted .nii.gz files (DWI and T1) to .nii using freesurfers mri_convert
2) Converted *.bval/*.bvec to B-matrix.txt file.
3) Used Flip/Permute dimensions of 3D/4D *.nii files with the default parameters.
4) Shuffled DWIs so that the b0 weighted volumes were in the beginning.
5) Calculated DTI *.mat file.
6) Corrected for subject motion & EC/EPI distortions
a) Registering to other data to do the EPI correction (non-rigid)
b) Defining the image type as FA
c) Setting the deformation axes as 1 0 0
d) Leaving all other settings as they are per default in ExploreDTI v4.8.6
I have 2 scans from the participant in question, because we conducted a longitudinal study with ~4 weeks between scans. Please find the 2 corrected FA images attached (exported from the trafo.mat file).
I have loaded both the T1 and the FA into FSLVIEW and what I took screenshots of is the colored T1 image and the grayscale FA map. In the scan1_brain_ok image, you see very little of the T1 under the FA. However in the scan2_brain_squashed image, you can see a ring of T1 around the whole FA map.
Any suggestions?
Thank you in advance!
stijn m
May 9, 2017, 3:43:29 AM5/9/17
to ExploreDTI
Dear Clara,
The first thing that comes to my mind is that you might have swapped the two T1 scans T=0 and T=4 weeks. How does your longitudinal design look like? Are you mapping the DWI to the T1 for both timepoints separately? Or maybe map the DWI at T=4weeks to the T=0 in order to make them more comparable?
Considering the processing steps you take; you can leave out the first step and just unzip the nifti's. It would be a nice feature if ExploreDTI could handle zipped files. Step 6b is maybe incorrect as you wrote that you register the DWI to T1 and not to the FA. Therefore the image type would be non-FA.
HTH
BW.
Stijn
The first thing that comes to my mind is that you might have swapped the two T1 scans T=0 and T=4 weeks. How does your longitudinal design look like? Are you mapping the DWI to the T1 for both timepoints separately? Or maybe map the DWI at T=4weeks to the T=0 in order to make them more comparable?
Considering the processing steps you take; you can leave out the first step and just unzip the nifti's. It would be a nice feature if ExploreDTI could handle zipped files. Step 6b is maybe incorrect as you wrote that you register the DWI to T1 and not to the FA. Therefore the image type would be non-FA.
HTH
BW.
Stijn
Clara
May 9, 2017, 4:20:15 AM5/9/17
to ExploreDTI
Dear Stijn,
Thank you for your quick reply!
We collected MP2RAGE and DWI data from 4 year children at T=0, T=4 weeks, and T=8 weeks. I am mapping the DWI to the T1 separately because I do not want to introduce any bias. If I register T=4 weeks DWI to T=0 T1, I think the deformation may be larger than the deformation applied when I register T=0 DWI to T=0 T1. This would be systematically treating the scans differently and might introduce a confound when I look at differences between the scans.
I don't think the problem is that I switched the T1s, because I have now run the correction several times, and each time I started from scratch.
Finally, I used FA in the contrast because on page 24 of the manual it is written that "for a T1, you could also try 'Avg(DWIs)' or 'FA'. Additionally, in this another post in the google group, Alexander Leemans said " 2. The default, "Non-DWI" works best if your reference image is a T2 image. For a T1, I would choose the "FA" or "Avg(DWIs)" contrast as it looks more similar to the T1." (https://groups.google.com/forum/#!searchin/e_dti/fa$20or$20non$20fa%7Csort:relevance/e_dti/eoVtRDqOZZg/tTwyHTVVdIwJ)
I think the strange thing is that I have done this with 60 other participants, but don't have such a problem with the exact same settings.
Kind regards,
Clara
Thank you for your quick reply!
We collected MP2RAGE and DWI data from 4 year children at T=0, T=4 weeks, and T=8 weeks. I am mapping the DWI to the T1 separately because I do not want to introduce any bias. If I register T=4 weeks DWI to T=0 T1, I think the deformation may be larger than the deformation applied when I register T=0 DWI to T=0 T1. This would be systematically treating the scans differently and might introduce a confound when I look at differences between the scans.
I don't think the problem is that I switched the T1s, because I have now run the correction several times, and each time I started from scratch.
Finally, I used FA in the contrast because on page 24 of the manual it is written that "for a T1, you could also try 'Avg(DWIs)' or 'FA'. Additionally, in this another post in the google group, Alexander Leemans said " 2. The default, "Non-DWI" works best if your reference image is a T2 image. For a T1, I would choose the "FA" or "Avg(DWIs)" contrast as it looks more similar to the T1." (https://groups.google.com/forum/#!searchin/e_dti/fa$20or$20non$20fa%7Csort:relevance/e_dti/eoVtRDqOZZg/tTwyHTVVdIwJ)
I think the strange thing is that I have done this with 60 other participants, but don't have such a problem with the exact same settings.
Kind regards,
Clara
stijn m
May 10, 2017, 7:54:00 AM5/10/17
to ExploreDTI
Dear Clara,
Preventing a bias towards one scanpoint could be a valid reason to map the DWI's to the T1 images separately.
It seems to depend on the SNR you get from the T1 images. Maybe something went wrong for this specific case?
BW.
Stijn
Preventing a bias towards one scanpoint could be a valid reason to map the DWI's to the T1 images separately.
It seems to depend on the SNR you get from the T1 images. Maybe something went wrong for this specific case?
BW.
Stijn
Clara
May 10, 2017, 9:12:07 AM5/10/17
to ExploreDTI
Dear Stijn,
It seems like the problem was the voxel sizes.
In the "Used Flip/Permute dimensions of 3D/4D *.nii files with the default parameters" step, where I flipped/permuted my DWI and T1 data, the voxel size of my diffusion data changed from 1.9 isotropic to 1.9 x 1.887408 x 1.887408.
There is the possibility of forcing the voxel size during this step. When I do this (force voxel size to 1.9 x 1.9 x 1.9) - the SM/EC/EPI corrected FA image is the same size as the T1, so this was the problem.
Kind regards,
Clara
It seems like the problem was the voxel sizes.
In the "Used Flip/Permute dimensions of 3D/4D *.nii files with the default parameters" step, where I flipped/permuted my DWI and T1 data, the voxel size of my diffusion data changed from 1.9 isotropic to 1.9 x 1.887408 x 1.887408.
There is the possibility of forcing the voxel size during this step. When I do this (force voxel size to 1.9 x 1.9 x 1.9) - the SM/EC/EPI corrected FA image is the same size as the T1, so this was the problem.
Kind regards,
Clara
Reply all
Reply to author
Forward
0 new messages