CSA normalisation


I have CSA results for a group of patients in the cervical and thoracic cord (not all have brain imaging). I wondered if you could recommend a technique for normalisation of these values using the cord imaging?

Is it possible to extract a length value from the straightened cord during registration (as this would then prevent curvatures etc from confounding the normalisation if you just used the raw number of slices)?

Any advice would be greatly appreciated!


Hi Romina,

Sorry for the delay. CSA normalization is a controversial topic. There is no clear consensus on how to properly normalize CSA across subjects.

Normalizing by cord length is possible in the straightened space, but you would need to have the full picture of the spinal cord, i.e., from the brainstem down to the lumbar cord (the cord typically stops at around L1-L2 in adults).

Hope that helps!

Hi Julien,

Not a problem. I don’t have a full image of the cord - the images I have are cervical (BS to C7) and then thoracic (T1 to around L1 in most subjects).

How would you recommend I address the issue with what I have? Particularly if wanting to publish.



As I said, normalization of CSA is still an ongoing debate. Many publications out there have not normalized CSA, so I don’t think this is a major problem to report unnormalized CSA values.

Sometimes, by attempting to normalize, you might end up introducing more noise (because the normalization factor is also affected by measurement noise), so while you might gain precision at one place, you also decrease precision at another place…


Great, thank you very much for this! I really appreciate the advice and the help!