1
$\begingroup$

I am research in brain segmentation task that classify a brain in to 3 classes : White Matter, Gray Matter and CSF. To validate my method, I used Dice Coefficient that compare my result and ground truth result. However, I am confusing about ground truth result. As I found that ground truth brain of brainweb service has two models: fuzzy and discrete models.

A fuzzy phantom will specify a mixture of tissue classes for each voxel (eg. a voxel might contain 90% grey matter and 10% white matter). A fuzzy phantom was used to describe the tissue within each voxel during the simulation process.

A discrete phantom specifies the tissue type with the largest proportion in each voxel. A discrete phantom was not used in the simulation process. However a user might use this phantom when comparing simulated data against the underlying tissue types (eg. a user might have a classification algorithm which classifies each voxel into one specific tissue type).

They suggest that "If your method attempts to estimate the partial volume (that is, the tissue fractions in each voxel) then the fuzzy model is probably what you want. Otherwise, the discrete model will probably do..."

From that information, Which model can I used for segmentation task?

As my knowledge, I think that discrete model must be used. How about you?

$\endgroup$
1
$\begingroup$

Your question practically answers itself in the "If your method attempts to estimate the partial volume then the fuzzy model is probably what you want. Otherwise, the discrete model will probably do..." quote.

So: if you're only going to segment each voxel into one of white/gray/CSF then yes, the discrete one seems the appropriate one to use.

$\endgroup$
  • $\begingroup$ Thanks timday. It is very useful for me. I have one small question that is in discrete model, after skull removing, we have four classes White/Gray/CSF and Glial Matter (label=8) . However, my goal is only three classes white-matter/gray-matter/CSF. Is it true if I combine gray mater (label=2) and Glial Matter (label=8) as a class. $\endgroup$ – John Aug 5 '15 at 7:46
  • 1
    $\begingroup$ Sorry don't know enough about the anatomy; I know there are glial cells which occur in both grey and white matter... but I'm not sure what people are actually segmenting when they talk about segmenting "glial matter" as something separate from grey or white, or whether it's considered closer to grey or white in some way. $\endgroup$ – timday Aug 5 '15 at 20:40
  • 1
    $\begingroup$ Most segmentation algorithms, applied to clinical MR images, would not include this class and so would treat this tissue as white matter and WM partial voluming with CSF. Many users of the Brainweb simulations also ignore the glial matter and focus only on the grey matter, white matter and CSF, using the phantoms to identify these tissues. However, this procedure introduces a significant difference between evaluation on Brainweb images and application to clinical MR: a significant amount of partial voluming between CSF and WM has been removed from the former in comparison to the latter $\endgroup$ – John Aug 7 '15 at 4:40

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.