How to stitch together different spinal cord segments' DTI images (e.g., cervical-thoracic, thoracic-lumbar)?

[willimon]

Hello,

I am currently performing spinal cord DTI analysis using the Spinal Cord Toolbox (SCT) and have encountered an issue. I have DTI images from different spinal cord segments, such as cervical-thoracic and thoracic-lumbar, and I would like to stitch them together into a single image for further analysis.

I understand that the sct_image tool has a -stitch command for image stitching, but I am unsure whether it can be used to stitch together images from different spinal cord segments (e.g., cervical-thoracic, thoracic-lumbar). If possible, could you please guide me on how to correctly perform this operation? Are there any specific commands or steps to achieve this?

Thank you for your help!

[joshuacwnewton]

Hi @willimon,

Just to make sure I’m fully understanding your use-case: Do the two segments overlap in the thoracic region for your subjects?

With that in mind, there are some caveats to mention:

• The original paper behind sct_image -stitch was intended for stitching together full-body scans. That said, sct_image -stitch has been applied to partial spinal scans by users in the past.
• The original paper also used T2w images, whereas you are working with DTI images that may have thicker, axially-acquired slices. Additionally, there may also be concerns about the effect of interpolation on your data.

I would recommend reading this reference thread on the forum. Julien authored a reply that seems very relevant for your use-case:

• Stitching and segmentation

@jcohenadad

I would not recommend stitching the raw (ie: DWI data), because the interpolation would impact noise distribution and model fitting. Instead, stitching should be done on the computed metrics (FA, etc.). The decision to stitch also depends on what the analysis will consist of. Eg: if the goal is to extract DTI metrics in specific tracts, then there is no need for stitching. Another thing to consider, is that the stitching algorithm assumes the slabs overlap sufficiently, and the structure overlaping the two slabs is perfectly registered. This assumptions is likely violated in the context of EPI scans, which are often hampered by susceptibility-related distortions.

The discussion as a whole is likely worth reading.

Julien is currently away from the lab due to summer break, but if you need further guidance, please let us know and we will try to respond as promptly as we can.

Kind regards,
Joshua