Normalisation of a postmortem dataset


I was wondering whether there was a way to obtain spinal levels automatically in the case of a postmortem dataset. Ideally we would love to normalize the dataset to PAM50, but we don’t have vertebrae — since it’s a postmortem acquisition.
Is there a workaround for this type of normalization?
Would it be possible to get this info without figuring out and counting manually the nerve roots?

Thank you in advance!

Hello @iricchi, thank you for reaching out.

Unfortunately there is currently no automated method to identify the spinal nerve rootlets, so you will need to do it by hand. Working in the straightened space is much easier for identifying the rootlets (especially when looking at the coronal view). Maybe @sandrinebedard has additional tips for you.

Once you’ve labeled each spinal level, you can register to the PAM50 template using the -lspinal flag. Also see this similar post Registration to T1 ex-vivo Template ERROR.


Hello @iricchi,

To identify the spinal nerve roots, you can also apply denoising which helps visualizing the spinal nerve roots.

Here is a procedure to identify nerve roots: GitHub - sct-pipeline/pmj-based-csa: CSA measure based on distance from pontomedullary junction (PMJ)

Here is the pipeline used for straightening and creating the labels: pmj-based-csa/ at main · sct-pipeline/pmj-based-csa · GitHub