ENHANCE-PET / MOOSE

MOOSE (Multi-organ objective segmentation) a data-centric AI solution that generates multilabel organ segmentations to facilitate systemic TB whole-person research.The pipeline is based on nn-UNet and has the capability to segment 120 unique tissue classes from a whole-body 18F-FDG PET/CT image.
https://enhance.pet
GNU General Public License v3.0
196 stars 28 forks source link

Bug: Small Intestine/Necrotic tumor tissue missclassified as bladder #21

Closed KCHK1234 closed 1 year ago

KCHK1234 commented 2 years ago

Fluid dense tissue surrounding the bladder often delineated as bladder while bladder delineation often only partial, probably hard to fix due to many instances of gravity dependend ascites in malignancies and cervical cancers as cohort. All instances recorded

josefyu commented 2 years ago

Does all "instances recorded" mean the error appears consequently in all images you have seen?

KCHK1234 commented 2 years ago

No only that I made a list of all instances in which i noticed it.

LalithShiyam commented 2 years ago

@KCHK1234 and @josefyu, the more I think of it, the bladder seems to be particularly hard to segment using low-dose CT. Why don't we do this in the PET? Also, why do we need bladder again? Is there any indication where bladder segmentation is important? My gut feeling says, if we could segment the bladder accurately from PET, we can possibly obviate the mix up between the tumour and the bladder... Other than this I don't see any use for bladder segmentation. Thoughts?

KCHK1234 commented 2 years ago

Yes bladder segmentation would work great in FDG and PSMA, but you will get problems with non renal excretion. For the organ axis stuff bladder is prob. not interesting anyways, but if you wanna push this into theranostic/dosimetry, bladder is sometimes the critical organ (Dose limiting) and therefore should be segmented properly (if I remember correctly - gotta check that in a book).

LalithShiyam commented 2 years ago

@josefyu : I agree with @KCHK1234, lets make sure we segment the bladder from the PET and not the CT. And frankly moose is just for FDG now. So this shouldn't be a problem.

josefyu commented 2 years ago

so right now, I am segmenting FALCON (siemens dataset 4 F, 4 M, 14 sequences), and here I use PET to delineate the bladder.