Analysis: Fine-tune presegmentations in SMS Vision datasets (8 Healthy, 8 Cancer)
LalithShiyam opened this issue · 8 comments
For this I need a briefing, what does the finetuning presegmentation stands for. Does it mean, SMS were MOOSEd and I shall check, if they have errors?
What shall we do with the cancer patients, do we label tumor as an extra segment?
The workflow is the following:
- @LalithShiyam sends in the paired-PT and paired-labels for the healthy volunteers for now (along with the paired-CT of course).
- @LalithShiyam to prepare the paired-PT and paired-labels for the lung cancer patients.
- @josefyu corrects the segmentations as accurately as possible
Data stored on
HD006/FALCON/Analysis
Patient cohort reviewed two times,
all labels modified for: brain, bladder
labels sometime needed to be modified: kidneys, liver, spleen
mostly untouched - lung
NRRD Segmentation for SMS Vision - Healthy
Images are on NAS
SMS-vision-healthy
Hi @josef, please don't upload images here. Or the IP address. Just mention that they are present in the HDD/ELN :D once the repo goes open source everyone can see these things. Not the best way.
removed files on github / and IP
Daria has run moose on ultra low dose vision dataset.
the standard dose dataset is dynamic and MOOSE couldn't process them.
SMS dataset with 8 patients cancer also segmented.
data structure prepared as Sebi wished, and copied to LUCAPET network drive