Consider quantifying g-factor within a mask, or overlaying an anatomical image

[dpapp86]

Related to #62 and #58 and partially to #56

G-factor maps are usually shown either for simple phantoms, or for "full" anatomy (a brain, an abdominal slice, etc), which makes theme easy to interpret, see examples in #62 and #58.

By contrast, our maps are taken of an area that is mostly "empty" of signal, and will thus not produce nice-looking g-factor maps with a clear border showing foldovers and areas of high g-factor.

To improve readiblity, I suggest we either agree on a mask in which the g-factor can be quantified (this could be some extended SC mask, since we do not care that much about g-factor in the ribcage), or an anatomical overlay be provided, so the reader can easily see what is important.

[evaalonsoortiz]

@jcohenadad As discussed at ISMRM: we could use the model that was used to design Spinoza to extract a spinal cord ROI. Could you provide me with the images of that model?

[jcohenadad]

we could use the model that was used to design Spinoza to extract a spinal cord ROI. Could you provide me with the images of that model?

Discussed here: https://github.com/neuropoly/phantom-spinoza/issues/5

[jcohenadad]

As mentioned in https://github.com/neuropoly/phantom-spinoza/issues/5#issuecomment-2150307589 we should create a spinal cord mask programatically. @evaalonsoortiz I can take care of it and add the code in the Notebook if you 'thumb up' this comment.

[evaalonsoortiz]

Update:

I'm now using a spinal cord mask to compute B1+ and SNR in the phantom along the "fake" spinal cord (see: https://github.com/spinal-cord-7t/coil-qc-code/issues/64#issuecomment-2167074164)

This solution requires labels for the TFL, DREAM, and SNR images so that a spinal cord mask could be defined for each one of those images and so that scans could be co-registered across sites.

If we want to use the spinal cord mask to quantify the average and maximum g-factor, I would need to be able to visually identify the registration label (as was done for TFL, DREAM and SNR). This is not so easy to identify in a g-factor map (see images below). So, my solution is to define a square ROI in the center of the image and compute the average and max g-factor there (this is commonly done in publications from what I've seen).

Phantom:

In-vivo:

If I can get a thumbs up to approve this solution, I'll close this issue.