I'm trying to measure the DICE coefficient of brain segmentations by a highresnet model. For my ground truth, I'm using FreeSurfer segmentations. There is some overlap between the regions used by each, but there are a lot of FreeSurfer regions that are not contained in GIFNiftyNet.ctbl. That is, they don't have the exact same name. I imagine that each of the FreeSurfer regions is a part of a region listed in GIFNiftyNet.

Does anyone know of a mapping from FreeSurfer region numbers to GIFNiftyNet numbers?

Also, has there been any attempt to standardize the numbering and names given to parts of the brain by various segmentation programs? How are DICE correlations for brain segmentations usually computed? That is, what is used as the ground truth?

  • 1
    $\begingroup$ You'll probably have to put them in some shared space and compare the mappings yourself. They're unlikely to be 1:1, in that Region 1 from A might map into Region 1, 2 and 3 from B, while Region 2 from A also maps partially into Region 1 and 3 from B, etc. $\endgroup$
    – Bryan Krause
    Apr 9, 2021 at 16:21


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.