cci-bonn / hd-glio-auto Goto Github PK

Hi everyone!

First of all thanks for this tool. Although the prerequisites were difficult to meet, now I am able to use it.

However, for some patients, the script get stuck after betting and coregistration (I can see the output such as '*bet_reg.nii' or '*bet_reg_norm.nii').
All of the input images are correct, as far as I know.

Do you have any idea of why this happened?

Thanks in advance,

Nicolo

Hi hd-glio-auto developers,
Dear Jens,

I'm trying to use hd-glio-auto on a recent nvidia gpu.

I get the following error when running the container:

>                 jenspetersen/hd-glio-auto
/usr/local/lib/python3.6/dist-packages/torch/cuda/__init__.py:125: UserWarning: 
NVIDIA RTX A5000 with CUDA capability sm_86 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_70 sm_75.
If you want to use the NVIDIA RTX A5000 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

I had something similar with hd-bet, but upgrading my Torch to a more recent version fixed it. Pythorch with cuda toolkit 11.3 e.g. by running:

pip3 install torch==1.10.1+cu113 torchvision==0.11.2+cu113 torchaudio==0.10.1+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html

Now I would like to make my own docker container replacing the first line in the Dockerfile with:

FROM nvidia/cuda:11.4.2-cudnn8-runtime-ubuntu20.04

Should do the trick, I think.

However the docker build fails, even with the orginal Dockerfile file. On what platform did you build the docker container?

Any suggestions would be highly appriciated.

Kind regards,
Stefan.

I noticed that the model has been trained to segment Necrosis as well. In the run.py file, I observed some commented lines that seem to be related to Necrosis. I was wondering if you could provide more clarification on this matter.

Thank you in advance,

Nicolo

Thank you for sharing your incredible work for free.

I want to extract tumor segmentation mask from brain MRI images.
The required mask type for me is not only contrast-enhanced and peri-tumoral but also necrosis.

In run.py script as given, the parameter 'necrosis_to_background' is set to True and
the result segmentation mask doesn't have necrosis class due to the default value.

How can I access the necrosis mask result on the provided docker image?
Thank you.

AssertionError:
Found no NVIDIA driver on your system. Please check that you
have an NVIDIA GPU and installed a driver from
http://www.nvidia.com/Download/index.aspx
Using contrast CT1 as reference
Traceback (most recent call last):
File "scripts/run.py", line 505, in
not args.no_permissions
File "scripts/run.py", line 280, in run
output1 = subp.check_output(["hd-bet", "-i", file_, "-device", "0"])
File "/usr/lib/python3.6/subprocess.py", line 356, in check_output
**kwargs).stdout
File "/usr/lib/python3.6/subprocess.py", line 438, in run
output=stdout, stderr=stderr)
subprocess.CalledProcessError: Command '['hd-bet', '-i', '/output/T1_r2s.nii.gz', '-device', '0']' returned non-zero exit status 1.

Thx for making available hd-glio-auto!

I observed a possible problem with the reported values in volumes.txt.

Seems like the reported volume in mm3, equals the number of voxels in the segmentation, but not the actual volume in mm3!
If voxel sizes are isotropic 1x1x1 mm these are correct.
However, when using a different resolution the output seems to be wrong.

An fslstats -V of the different segmentations shows this.

Thank you for sharing your nice works.

While running on my docker environment, the error message below occurs only for large size images. (>100 MB, 1024x1024 px)

Image Exception : #22 :: Failed to read volume /output/FLAIR
Error : Error: short read, file may be truncated
terminate called after throwing an instance of 'std::runtime_error'
  what():  Failed to read volume /output/FLAIR
Error : Error: short read, file may be truncated
Image Exception : #22 :: Failed to read volume /output/CT1
Error : Error: short read, file may be truncated
terminate called after throwing an instance of 'std::runtime_error'
  what():  Failed to read volume /output/CT1
Error : Error: short read, file may be truncated
Aborted (core dumped)
Aborted (core dumped)

To decrease the input file size, I resampled the image to downscale the volume below 20 MB but it failed.
Another trial was to increase the memory size of docker container by setting --memory 10G but it also not performed well.

I think the error occurs because the file size is too large for FSL to handle with.
Do you know any solution to tackle this problem?

Best,
Dongmin Choi.