I have thoracic cord T2 images acquired at 3T (starting at T1). I would like to calculate CSA values in these images. Are there any recommended adjustments to the code (and would I label C7/T1 as 8 and initiating the vertebral labeling from that point)?
Indeed, you can initiate labeling at C7/T1 and then run the labeling, example below:
sct_label_utils -i t2.nii.gz -create-viewer 8 -o label_disc_C7T1.nii.gz
sct_label_vertebrae -i t2.nii.gz -s t2_seg.nii.gz -initlabel label_disc_C7T1.nii.gz
Let us know if you need further help,
Thank you very much!
As a follow up question - most of my thoracic cord images segment well with deepseg; however in one or two of them the segmentation is not continuous at the lower extremity (around T12 - as would be expected depending on where the conus is in each participant). I am currently using manual vertebral labeling and so when registering to template those problem areas lower down do not appear on the registration image as I do not label that far down. I then go on to warp and compute the csa.
Does the poor segmentation lower down (which I am not registering to the template) create any problems with my results that I might be unaware of?
If you are not focusing on the lower cord, then you don’t need to fix the segmentation there and you can just go ahead with the rest of the analysis.
Let us know how it goes
Great, thanks Julien!
Is there a way to include the conus in the analysis?
It depends what type of analysis you wish to do. One possibility would be to input the conus segmentation (that you probably need to manually correct, in case the segmentation didn’t do a good job) inside
sct_process_segmentation, in order to compute CSA of the conus, slice-by-slice.
If you wish to compute qMR metrics (e.g., MTR, etc.), you could also use the conus mask along with your metric of interest with the function