Alignment of skull-stripped T1s on hybrid DTI/pCASL MRI datasets

AFNI Message Board
1

Greetings AFNI experts,

I have an alignment related issue that you might be able to help with:

I have skull-stripped T1 anatomical datasets that I need to align on top of hybrid DTI/pCASL datasets. Both the T1 and DTI/pCASL datasets have been collected in the same scan session:

T1: 1mm isotropic voxels (256x256x256 voxels)
https://i.postimg.cc/2SZWMvn8/T1.png

Hybrid DTI/pCASL: 3.5mmx3.5mmx8.0mm (64x63x12 voxels) The one shown is an M0 image,
https://i.postimg.cc/9Xdk980C/DTI-p-CASL.png

I can get a relatively, maybe sort of OK alignment if I do the following:

  1. Resample the Hybrid DTI/pCASL images to the T1 resolution
    3dresample -master ${s}_freesurfer_brain.nii.gz -prefix IMG_PLD2000_M0_RS.nii.gz -input IMG_PLD2000_M0.nii

  2. Align using align_epi_anat.py and pretending the DTI/pCASL dataset is EPI
    align_epi_anat.py -anat ${s}_freesurfer_brain.nii.gz -anat_has_skull no -epi IMG_PLD2000_M0_RS.nii.gz
    -epi_strip None -anat2epi -epi_base 0 -partial_coverage -volreg off -tshift off -suffix _AL_m0_RS.nii.gz

The output is depicted in the animated gif below (image shows the TI overlaid on top of the DTI/pCASL image after alignment):

https://i.postimg.cc/tJrPGGqh/Alignment.gif

The alignment is not totally horrible but I was wondering if there is anyway this can be improved. I can share my data if this will help,

Thank you in advance for your help,

Valentinos

2

That looks okay. It’s pretty rough in the DTI/pCASL dataset. You might try visualization with edges. You could try the nmi cost function to see if that comes out with something better.

3

Greetings,

Thank you for the quick response. The nmi cost function in combination with -big_move worked significantly better. Still not quite there but this might be the best we can do given the quality of the DTI/pCASL dataset. Thank you for your help,

Valentinos