Seeking sct_register_to_template suggestions

MRI Processing & SCT
1

Hi all,

I am looking for suggestions to improve my SCT sct_register_to_template results. I have attached the QC folder for reference. Despite trying several options, the registration is still somewhat suboptimal. One specific issue I noticed is that the registered template (template2anat.nii.gz) often appears larger than my subject T2 (t2.nii.gz) spinal cord.

Here’s what I have tried so far:

Using vertebral labels 3–7

  1. Command:
sct_register_to_template -i t2.nii.gz -s t2_seg.nii.gz -l t2_labels_vert.nii.gz \
  -qc qc_singleSubj -c t2 \
  -param step=1,type=seg,algo=centermassrot:step=2,type=seg,metric=CC,algo=bsplinesyn,slicewise=1,iter=3:step=3,type=im,metric=CC,algo=syn,gradStep=0.2,smoothWarp=1.0,slicewise=1,iter=2
  1. Command (default parameters):
sct_register_to_template -i t2.nii.gz -s t2_seg.nii.gz -l t2_labels_vert.nii.gz -c t2 -qc qc_singleSubj
  1. Command (slightly modified parameters):
sct_register_to_template -i t2.nii.gz -s t2_seg.nii.gz -l t2_labels_vert.nii.gz \
  -qc qc_singleSubj -c t2 \
  -param step=1,type=seg,algo=centermassrot:step=2,type=seg,metric=CC,algo=bsplinesyn,slicewise=1,iter=3:step=3,type=im,metric=CC,algo=syn,slicewise=1,iter=2
  1. Command (more iterations and adjusted gradStep / smoothWarp):
sct_register_to_template -i t2.nii.gz -s t2_seg.nii.gz -l t2_labels_vert.nii.gz \
  -qc qc_singleSubj -c t2 \
  -param step=1,type=seg,algo=centermassrot:step=2,type=seg,metric=CC,algo=bsplinesyn,slicewise=1,iter=5,gradStep=0.25,smoothWarp=1.0:step=3,type=im,metric=CC,algo=syn,slicewise=1,iter=3,gradStep=0.15,smoothWarp=1.2

Using all disc labels

  1. Command (default parameters):
sct_register_to_template -i t2.nii.gz -s t2_seg.nii.gz -ldisc t2_seg_labeled_discs.nii.gz -c t2 -qc qc_singleSubj
  1. Command (complex parameters):
sct_register_to_template -i t2.nii.gz -s t2_seg.nii.gz -ldisc t2_seg_labeled_discs.nii.gz \
  -qc qc_singleSubj -c t2 \
  -param step=1,type=seg,algo=centermassrot:step=2,type=seg,metric=CC,algo=bsplinesyn,slicewise=1,iter=5,gradStep=0.25,smoothWarp=1.0:step=3,type=im,metric=CC,algo=syn,slicewise=1,iter=3,gradStep=0.15,smoothWarp=1.2

Additional information

  • I used SCT v6.5 mainly because the segmentation tends to give slightly larger cord masks than v7.1, but I have tried both versions.
  • Voxel size: 0.8 × 0.8 × 0.8 mm
  • Cord coverage: C1–T1

Questions

  1. How good can sct_register_to_template realistically get for high-resolution isotropic T2 data?
  2. Are there any tips or best practices to assess registration quality quantitatively and visually?
  3. Do you have suggestions for parameter adjustments or preprocessing steps that could improve template alignment and prevent AP cord expansion?

Thanks in advance for any guidance!
Best,
Jia
qc_singleSubj.zip (4.8 MB)

2

Thank you for your question! We would be happy to help. :slight_smile:

Perhaps I am missing something, but the 3rd sct_register_to_template entry in the QC report appears to address the enlargement issues in the SC mask:

sct_register_to_template -i t2.nii.gz -s t2_seg.nii.gz -l t2_labels_vert.nii.gz -qc qc_singleSubj -c t2 -param step=1,type=seg,algo=centermassrot:step=2,type=seg,metric=CC,algo=bsplinesyn,slicewise=1,iter=3:step=3,type=im,metric=CC,algo=syn,slicewise=1,iter=2

ezgif.com-animated-gif-maker

I do notice that there is still enlargement in the canal (this kind of creates an “optical illusion” of enlargement when flipping back and forth). But, just to clarify, are you aiming to address the canal enlargement too?

I will tag my colleagues who are more experienced with the registration parameters and their customization, and we will get back to you shortly. :slight_smile:

Kind regards,
Joshua