Dear MRtrix3 Experts,
Recently, I have been checking my dataset analysed by MRtrix3. I found in the part of Anatomically-Constrained Tractography (ACT) recommended ‘register T1-contrast anatomical image to the diffusion image series before any further processing of the T1 image is performed to avoid reorienting the diffusion gradient table’. However, due to analyzing fMRI first and not noticing this tutorial, there was a subsequent analysis of the T1 image. So I registered the DTI to T1 space using the command below, and checked the registration result.
flirt -in ${output_path_CSD}/mean_b0.nii.gz \
-ref ${output_path_CSD}/brain.nii.gz \
-dof 6 -omat ${output_path_CSD}/diff2struct_fsl_brain.mat \
-cost normmi
transformconvert ${output_path_CSD}/diff2struct_fsl_brain.mat \
${output_path_CSD}/mean_b0.nii.gz ${output_path_CSD}/brain.nii.gz \
flirt_import ${output_path_CSD}/diff2struct_mrtrix_brain.txt
mrtransform -linear ${output_path_CSD}/diff2struct_mrtrix_brain.txt \
${output_path_CSD}/${subjID}_preproc_biascorrect.mif \
${output_path_CSD}/${subjID}_preproc_biascorrect_brain_space.mif
I think the registration results are not bad. And I found the diffusion gradient table also changed compared with unregistered DWI images. Like this,
This is a gradient table that has been registered in T1 space.
This is a gradient table that is unregistered.
So I’m assuming that maybe MRtrix3 might have done a proper redirection of the gradient table based on the transformation matrix. I’m not sure if this assumption is correct, and I’d appreciate it if someone could tell me how to check it.
Furthermore, I generated separate FODs using a registered DWI, and an unregistered DWI using the same pipeline respectively, and normalized them. Like this,
docker run --rm \
-v ${output_path_CSD}:/data \
kaitj/mrtrix3tissue:v5.2.9 \
ss3t_csd_beta1 -force /data/${subjID}_preproc_biascorrect.mif \
/data/wm_response.txt /data/${subjID}_wmfod.mif \
/data/gm_response.txt /data/${subjID}_gmfod.mif \
/data/csf_response.txt /data/${subjID}_csffod.mif \
-mask /data/mask.mif
mrconvert -coord 3 0 ${output_path_CSD}/${subjID}_wmfod.mif - | \
mrcat -force ${output_path_CSD}/${subjID}_csffod.mif \
${output_path_CSD}/${subjID}_gmfod.mif - ${output_path_CSD}/vf.mif
mtnormalise -force -mask ${output_path_CSD}/mask.mif \
${output_path_CSD}/${subjID}_wmfod.mif ${output_path_CSD}/${subjID}_wmfod_normal.mif \
${output_path_CSD}/${subjID}_gmfod.mif ${output_path_CSD}/${subjID}_gmfod_normal.mif \
${output_path_CSD}/${subjID}_csffod.mif ${output_path_CSD}/${subjID}_csffod_normal.mif
I thought I should get the same FOD content between the registered and unregistered, but I found that didn’t turn out to be the case. Simply, I found that unregistered DWI images acquire bigger FOD values in the corpus callosum(same voxel index). If I understand the FOD correctly, a larger FOD value means that the fiber bundles here are more directional (Please let me know if anything is incorrect, I’m just a newbie to DTI!). I would like to know what is the cause of this discrepancy? Does this mean that I should start all the analysis again, i.e., fMRI as well as T1 analysis after transforming T1 to DWI space.
Thank you so much for receiving any insights or comments you may have, any input you have will help me immensely.
Yufeng