Jakob Wasserthal

Results 222 comments of Jakob Wasserthal

Extending TotalSegmentator to MRI is on our agenda, but it will be more in the future. This will take at least another year.

v1 was downloading the weights from zenodo which was slow. v2 is now downloading the weights from github. Not sure why this is blocked by your hospital IT.

The weights are only downloaded once. But if your download failed, then it does not find any weights on the next run and tries to download them again.

Thanks for mentioning this. TotalSegmentator depends on rt-utils for generating RTstructs. It does not make sense to implement my own logic around that. So I have to wait for them...

Which totalsegmentator version are you using? Can you update to the newest version?

Using gpu:1 is only possible in the newest master. In the latest pip version only `--device gpu` is possible.

When you install the newest pip package the python api will also use the new version. So I guess you still have the old totalsegmentator installted.

Calling the remaining commented code does not work. It is deactivated for a good reason ;). I first have to train a nnunet v2 on the liver vessel data. This...

The vertebrae_body mask is intended to be used to refine the individual vertebrae masks. Just take the intersection of both for each vertebrae. Then you have what you need.