Add application: NLmCED

[Palollo]

nlmced

Description

The NLmCED (Non Local mean Coherence Enhancing Diffusion) filter is a denoising tool aimed to reduce Rician Noise in 3D MR images. It is a combination between two filters such as the Non-Local mean filter and the Anisotropic Diffusion tensor method with an estimator of Rician noise. The NLmCED filter can benefit from the noise reduction capabilities of the NLM filter while maintaining the edge-preserving characteristics of anisotropic diffusion.

Usage

jobman submit -i nlmced -- <input_dir> <output_dir> [iter] [rho] [alpha] [wind]

NLmCED filter is an iterative filter with some parameters to be set:

• input_dir: the input path of a dataset directory.
• output_dir: the output directory where the results will be saved.
• iter (optional): iteration number (default value: 1).
• rho (optional): (ρ) a standard deviation of the Gaussian kernel for the creation of the structure tensor (default value: 0.01).
• alpha (optional): (α) a single value to control the diffusion tensor matrix (default value: 0.01).

Example:

  jobman submit -i nlmced -- ~/datasets/87f3be56-4725-45c3-9baa-d338de530f73/ ~/persistent-home/nlmced-results/ 1 0.01 0.01

Note the output directory path should be any within the persistent-home, which is shared between all desktops and jobs created by the user,
otherwise the results will be lost after the end of the job.

Authors

Feriel Romdhane
Institute of Biostructures and Bioimaging (IBB) of the National Research Council (CNR)

Contact info

ferielromdhane@yahoo.fr
https://github.com/FerielRamdhane/Denoising-CEST-MRI/issues

URL

Documentation:
https://docs.google.com/document/d/1kxN0Jrn6xukAi6kVSmBdo1GmfkXNZ-5B/edit?usp=sharing&ouid=114076787312989347088&rtpof=true&sd=true

Source code repository:
https://github.com/FerielRamdhane/Denoising-CEST-MRI/tree/main/NLmCED_Filter

Private* dockerfile repository:
https://github.com/FerielRamdhane/NLmCED-Filter

• You will see error 404 if you don't have permissions to access.

License

https://github.com/FerielRamdhane/Denoising-CEST-MRI/blob/main/LICENSE

[Palollo]

@FerielRamdhane, please review the draft but particularly these points...

Usage

Please confirm the parameters accepted by the tool, those which the user can put after the "--" in the jobman submit command (they will be appended to the ENTRYPOINT of your image).
Indeed the current version of the image (v1.0) has no entrypoint defined, you should change the line
CMD ["python" "./nlmced.py"]
to ENTRYPOINT ["python" "./nlmced.py"]
Also you should add the INPUT_DIR and OUTPUT_DIR params, as currently seems they are fixed to /input and /output, but user should be able to specified them.
We added the "wind" parameter here as we see in the documentation but we are not sure it exists.

Authors

You can add there your organization if you want.

URL

Dockerfile repository:
https://github.com/FerielRamdhane/Denoising-CEST-MRI/tree/main/NLmCED_Filter

We don't see the Dockerfile in that repo, please replace it with the public or private repository where you put the Dockerfile and all the dependecies available for us to build the image.

[FerielRamdhane]

@Palollo Thank you for the review and for pointing out these sections.

Usage

An entrypoint.sh script has been added, and the Dockerfile has been updated to use: ENTRYPOINT ["/app/entrypoint.sh"]. This update introduces new input and output directory parameters.
Regarding the wind parameter, it has been removed, as it was already optimized in this version and is no longer required.

Authors

Euro-BioImaging · Institute of Biostructures and Bioimaging (IBB) of the National Research Council (CNR) · University of Tunis

URL

The Dockerfile and build files are available in this repository:
https://github.com/FerielRamdhane/NLmCED-Filter

I have added you as a collaborator. Please let me know if any other team members also require access so I can add them as well.

[Palollo]

Thank you @FerielRamdhane for your attention and to resolve almost all the points.
Have you tried the new entrypoint? It seems not working with the two new params.
I suggest few changes in this PR: https://github.com/FerielRamdhane/NLmCED-Filter/pull/1

And as recommendation in order to not install the full Matlab runtime but instead only the components you need, to minimize the size of image (currently 11GB!), it seems compiler.runtime.customInstaller can be used for that:
https://www.mathworks.com/help/compiler/compiler.runtime.custominstaller.html
Have you checked it out?

[Palollo]

Just one point more: now there are three levels of visibility for the batch applications (details here).
Which level should be assigned to this application?

[FerielRamdhane]

@Palollo Sorry for the delayed reply, and thank you for your suggestions!
Regarding the new entrypoint: it works fine on my side, but I’ll definitely test your proposed changes. Thanks for taking the time to look into it.
About the MATLAB runtime: I haven’t tried using compiler.runtime.customInstaller yet, but I’ll check it out and see how I can optimize the container size. I’ll provide updates later.
For the batch application visibility, I agree that Public makes the most sense for this application.
One last small request: could you please update the Author field to mention at least CNR? There’s no need to list all our three authors.
Thanks again for your helpful feedback!