Brain and spine segmentation in one go?

Hello experts!

I have a small project where it would be advantageous to have brain and spinal cord segmentations in one file as the output of a single procedure. 7T data, GRE scans with poor resolution (anatomical scans of B1/B0 mapping), whole brain coverage, spine coverage to C4-ish or a bit lower.

Is there a best practice for this?

I have been playing around with using BET and sct_deepseg_sc separately, then merging the two masks and manually painting in the missing areas, but this is quite laborious and may be suboptimal.

I know this goes a bit against the grain, and is probably and edge case, but if anyone has experience, I’d be glad to listen.

Hey Daniel!

There are various ways to do this.

The most straightforward way is to work in the MNI space, which has the ICB152 brain template and the PAM50 spinal cord template (all in the same space coordinate system).

Another way to do this is to take your brain anatomical scan, add zero padding below (eg: sct_image -i IMAGE -pad 0,0,30) and register your SC scans to this space using the qform matrix (eg: sct_register_multimodal -i IMSRC -d IMDEST -identity 1).

Of course the best decision depends on what are the data, resolution, coverage, etc. so without knowing the fine details I can only give you ‘high level’ ideas.