MHub / GC - Minor cleanup gc_lunglobes

[silvandeleemput]

Closes #69

This PR adds two minor cleanups to the gc_lunglobes model code:

• My e-mail address has been updated to my new one.
• The default.yml pipeline has been simplified by removing the unnecessary nifticonverter since now this is supported by default by the DSegConverter as well.

[LennyN95]

dicomseg_json_path: /app/models/xie2020_lobe_segmentation/config/dseg.json
We no longer need to use the JSON (actually, the file has been removed already);
we've that changes implemented on the patch-models branch, waiting for final testing before we merge into main (soon).

[silvandeleemput]

I have merged the main branch (which includes the patch-models branch changes) into this branch to resolve the conflicts, this should be good to add now.

[silvandeleemput]

Fixed the commit version of the linked model repo at the latest main branch commit found on 2023/09/13

[silvandeleemput]

@LennyN95 @denbonte What is the current planned line of action for this PR? It appears the meta.json file needs to be updated to the latest JSON schema. I guess this will be something that you guys can do best since you created the mata.json file for this model.

[LennyN95]

Thx for your status request @silvandeleemput:

• meta.json will be updated by us for all already implemented models synchronously with our server's backend.
• Please make sure the PR branch is up-to-date with the main branch (currently 41 commits behind MHubAI:main).
• The model is currently crashing when more than one GPU is available and this requires explicit GPU selection on machines with >1 GPUs installed on the host machine. This needs to be addressed, as all MHub models are expected to run with the --gpus all option.

[silvandeleemput]

Thx for your status request @silvandeleemput:

• meta.json will be updated by us for all already implemented models synchronously with our server's backend.
• Please make sure the PR branch is up-to-date with the main branch (currently 41 commits behind MHubAI:main).

That's fine, notify me once you have updated the meta.json

• The model is currently crashing when more than one GPU is available and this requires explicit GPU selection on machines with >1 GPUs installed on the host machine. This needs to be addressed, as all MHub models are expected to run with the --gpus all option.

I created an issue for this here: #69, please provide the error codes and I'll forward the issue to the repo owner so hopefully he can fix it in the repo. Otherwise, we can easily patch it.

[silvandeleemput]

• meta.json will be updated by us for all already implemented models synchronously with our server's backend.
• Please make sure the PR branch is up-to-date with the main branch (currently 41 commits behind MHubAI:main).
• The model is currently crashing when more than one GPU is available and this requires explicit GPU selection on machines with >1 GPUs installed on the host machine. This needs to be addressed, as all MHub models are expected to run with the --gpus all option.

PR branch has been brought up-to-date and the > 1 GPU issue (#69) has been addressed by referring to the new release of the algorithm which includes a fix.

[LennyN95]

@silvandeleemput I checked the Metadata, and we need some minor information to merge this PR.

• specify the dates for code and publication (all date formats need to be yyyy-mm-dd)
• specify true/false under summary.data.public
• change summary.model.training to supervised

[silvandeleemput]

@silvandeleemput I checked the Metadata, and we need some minor information to merge this PR.

• specify the dates for code and publication (all date formats need to be yyyy-mm-dd)
• specify true/false under summary.data.public
• change summary.model.training to supervised

It has been updated.

[LennyN95]

Thank you @sil, looks like false must not be placed in quotes or is recognized as a string instead of a boolean.

[LennyN95]

When you're at it @sil, can you also update the slice thickness (see https://github.com/MHubAI/coordination/issues/137)?

[silvandeleemput]

Thank you @sil, looks like false must not be placed in quotes or is recognized as a string instead of a boolean.

Of course, fixed!

When you're at it @sil, can you also update the slice thickness (see https://github.com/MHubAI/coordination/issues/137)?

The link for the mentioned issue seems to be broken. For now, I change the slice thickness to 0.75mm in line with the pixel spacing of the training data.

[LennyN95]

The link for the mentioned issue seems to be broken. For now, I change the slice thickness to 0.75mm in line with the pixel spacing of the training data.

Sorry, @silvandeleemput that is pointing to a private issue repo we used to keep track of things internally; I forgot it wasn't public. But changes are looking good, I'll merge and run the tests. Thanks!

[LennyN95]

@silvandeleemput the model didn't provide any output.

I checked what could've went wrong: you removed the NiftiConverter but specified source_segs: nifti:mod=seg.
Solutions are to either remove that line (the default value for DSegConverter.source_segs contains mha, nifti and nrrd) or specify mha as the file type.

I'll fix this and run the build pipeline again. Please always test-run models after any implementational changes ;)