Available datasets

[valosekj]

• canproco (n=450, HC, MS)
  • T2w (complete C, part Th), 0.8mm iso, SC seg available
  • Other contrasts are available but need to be co-registered - WIP
• spine-generic (n=267, HC)
  • T2w (complete C, part Th), 0.8mm iso, SC seg available
  • T1w (complete C, part Th), 1mm iso, SC seg available
  • T2star (C2-C5), 0.5x0.5x5mm, SC seg available
  • MT (C2-C5), 0.9x0.9x5mm, SC seg available
  • DWI (C2-C5) 0.9x0.9x5mm, SC seg available
  • coregistration has already been done within contrast-agnostic project
• basel-mp2rage (n=283, HC, MS)
  • MP2RAGE (brain, C1-C5), 1mm iso, SC seg available
• datasets/lumbar-epfl
  • Lumbar SC

[naga-karthik]

On discussing this with @jqmcginnis (Julian, from TUM, datasets/bavaria-quebec-spine-ms), he mentioned having GT centerline data (labelled by experts) for the Bavaria-quebec project. That could be one more dataset to train our model on.

EDIT: it's technically not just the GT centerlines but the dataset has spinal cord segmentations (along with lesion segmentations) so the centerlines could be extracted from the SC segmentations.