You can ask MONAILabel developers to help you port the old model into a MONAILabel app.
Thanks for the ping, @lassoan
@Pari_TF, have you considered using MONAI Label (GitHub - Project-MONAI/MONAILabel: MONAI Label is an intelligent open source image labeling and learning tool.)?
It is easy to get a pretrained model for lung segmentation using MONAI Label. Are you using a public dataset?
Thanks for your reply. No, I haven’t used MONAI Label. That’s great. I will try it. Thank you so much. Yes, I’m working on a public dataset.
I am using my own data to segment lung nodules. The training set has been using the annotation_lung_tumor model of Nvidia AIAA. Now that the test set is ready, I found that the above model is no longer available, which caused me a lot of trouble. Can you help me add the old model to MONAI Label, thanks.
Do you know which dataset they used to get that pretrained model? I think it is Task06_Lung (MSD - Google Drive) from the Medical Segmentation Decathlon but I’m not sure.
If that’s the dataset, you should be able to get a pretrained model fairly easy. Do you have a GPU-based PC?
It is indeed a model pre-trained with the Task06_Lung dataset, and we use boundary-point based segmentation of lung tumor on CT (AIAA for 3D Slicer). But now the model is not found in NVIDIA. Also, we have GPU-based PC.
I was going to try this out but I am consistently running into
monailabel apps --download --name deepedit --output apps Using PYTHONPATH=/home/rbumm: App deepedit => /usr/monailabel/sample-apps/deepedit not exists
pip install monailabel
This is with and w/o an elevated bash on Windows 10 Ubuntu WSL. Any ideas?
@rbumm there were recently some changes in folder structure. Please try out the monailabel-weekly instead (pip install monailabel-weekly)
Then download the radiology app as follows:
monailabel apps --download --name radiology --output apps
As an example you could download and use the spleen dataset:
monailabel datasets --download --name Task09_Spleen --output datasets
Then start the server using the Deepedit model:
monailabel start_server --app apps/radiology --studies datasets/Task09_Spleen/imagesTr --conf models deepedit
Here you can see more details: GitHub - Project-MONAI/MONAILabel: MONAI Label is an intelligent open source image labeling and learning tool.
Thanks, Andres, tried this already - pip uninstalled monailabel and pip installed monailabel-weekly, but there is a similar message:
monailabel apps --download --name radiology --output apps Using PYTHONPATH=/home/rbumm: App radiology => /usr/monailabel/sample-apps/radiology not exists
Probably a permission problem.
This is strange
Probably the weekly installation wasn’t clean. Please do a clean installation using a clean Python virtual environment.
Needed to install monailabel via github to make this working in Windows 10 WSL (see here).
My bad, I just realised the weekly version doesn’t install the candidate release 0.4.0
In order to use the radiology app you should either use directly the GitHub repo as you did or install directly the candidate release:
pip install monailabel==0.4.0rc3
Please let me know how that goes. Happy to help.