Choosing the appropriate image contrast per task

[jcohenadad]

Based on the data @plbenveniste and I received from the HC, we can already make decisions about what image contrast to choose for each task. Below is a proposal:

• SC segmentation (contrast-agnostic)
  • Csp Ax T2
  • Csp Sag 3D Fiesta
  • Csp Sag Cube T2 Focus
• vertebral labeling (TotalSpineSeg)
  • ORIG SAG T2 LOC UPPER DL
  • also consider other contrasts used for SC seg if they are "reliable enough"
• lesion segmentation (PL's model)
  • Sag T2 STIR
  • Ax T2
  • Ax MERGE

[plbenveniste]

From your comment, I understand that we should use the original file only for vertebral labeling. For the rest, we should use the post-processed file right ?
Also, for the lesion segmentation, do you want to include only cervical spine ? Or thoracic and lumbar as well (in this case, it would make sense to include thoracic and lumbar in the previous sections as well) ?

[jcohenadad]

From your comment, I understand that we should use the original file only for vertebral labeling. For the rest, we should use the post-processed file right ?

not necessarily-- we could go with the denoised one (if this is what you meant by "post-processed")--
Also, we should consider the other contrasts-- i've updated my initial post

Also, for the lesion segmentation, do you want to include only cervical spine ? Or thoracic and lumbar as well (in this case, it would make sense to include thoracic and lumbar in the previous sections as well) ?

let's start with cervical only, for the sake of simplicity-- I suspect there is little amount of lesions in the thoraco-lumbar-- but as we receive the data we can reconsider