Dear developers,

In 5/25 of my patients, the software segmented parts of the eye as necrotic tissue/ (non-) enhanced tumour tissue. Is there a way I can exclude this from the analysis?

I am using version 2.0.

Best wishes,
A. Dijkshoorn

Hi Mauricio,

I tried two different things for the resulting cases.
(1) perform skullstripping in FSL by using bet and choosing the best value for skull-stripping
(2) create a manual mask for the T2-scan, co-registered the T1/T1C/FLAIR scan to the T2-weighted scan and then use this mask to extract the brains of the coregistered T1, T1C and FLAIR scan. Due to the intensive manual labor of creating the mask we chose for the T2-weighted scan.

Of the 5 cases that were initially not segmented properly, 3 cases remained questionable by visual inspection. I added a picture of both strategies. Do you have any idea why the tumour was not recognised by the segmentation software? Could you help me with this (I am unfortunately on a strict deadline, so time is running out..)

Hi Aicha, this is then mostly due to a bad skull-stripping that left some non-brain areas inside the mask. I recommend to clean those areas and re-run the tool (deactivating the skull-stripping) to avoid those false positives. Alternatively, you could run another skull-stripping approach on those cases and then segment the resulting skull-stripped cases
Best
Mauricio

Hi Mauricio,

Thank you very much for your quick reply.

I have applied skull stripping before running the segmentation, but this did not entirely remove the cranial orbit. I added a picture for clarification.

Best wishes,
Aicha

Hi Aicha, thanks for your interest in BratumIA. The tool had the option to perform skull stripping so only brain tissues are considered for the segmentation. Did you try that?
Regards
Mauricio