0.5 release activities

[SachidanandAlle]

• [x] Upgrade dependency
• [x] Unit Test and Integration pass
• [x] Verify Radiology App for all models (infer, train on single/multi-gpu)
• [x] Verify Pathology App for all models (infer, train on single/multi-gpu)
• [x] Verify Endscopy App for all models (infer, train on single/multi-gpu)
• [x] Verify Bundle App for all models (infer, train on single/multi-gpu)
• [x] Verify client 3D Slicer
• [x] Verify client OHIF (next sample, infer, train)
• [x] Verify client QuPath
• [x] Verify client DSA
• [x] Verify Docker
• [x] Scribbles (OHIF + 3D Slicer) (best works in 3D Slicer due to ROI)
• [x] Verify all documents
• [x] Release candidate

[SachidanandAlle]

First step: https://github.com/Project-MONAI/MONAILabel/pull/979

[SachidanandAlle]

Depends on: https://github.com/Project-MONAI/MONAILabel/pull/981

[SachidanandAlle]

Verified Radiology App for all models (infer, train on single/multi-gpu) - done for segmentation, segmentation_spleen, deepgrow 2d+3d+pipeline(infer), deepedit models

Verified vertebra for infer only @diazandr3s please verify once again.. and may be train if that's applicable. Also you need to add one section for this here: https://github.com/Project-MONAI/MONAILabel/tree/main/sample-apps/radiology#model-overview (may be you can combine all models under one section..) and mention about pipeline

[SachidanandAlle]

Upgrade to rc2: https://github.com/Project-MONAI/MONAILabel/pull/983

[tangy5]

Bundle app Verfications:

monai version == 1.0.0rc2

Setup

• [x] Download the monaibundel app and all model-zoo models. Verification of current Model-Zoo models 3D Bundles (3D Slicer)
• [x] Train/Inference 1: [spleen_ct_segmentation] 2: [swin_unetr_btcv_segmentation]
• [x] Inference: 1: [renalStructures_UNEST_segmentation] 2: [wholeBrainSeg_Large_UNEST_segmentation]
• [x] Train/Inference 1: [brats_mri_segmentation] 2: [spleen_deepedit_annotation]
• [x] Train/Inference: 1: [pancreas_ct_dints_segmentation] 2: [prostate_mri_anatomy] 2D Bundles
• [x] Train/Inference: 1: [pathology_tumor_detection]
• [x] Train/Inference: 1: [valve_landmarks] 2: [ventricular_short_axis_3label] 3: [mednist_gan]

Notes:

• Bundle [swin_unetr_btcv_segmentation] : Training with SwinUNETR bundle for multi-organ segmentation needs at least 32G GPU with current default parameters (batch size of 2 and number_sample=4), it will cause OOM for most users. Might need to set default batch size=1, number_sample=2 in the bundle for smooth run with MONAI Label.
• For customized scripts bundle, need to specify script path first, or it will fail instantiation custom class, then MONAI Label server will terminate. Need to remind users in the MONAI Label Bundle App, doc/readme. See the bundle doc at https://github.com/Project-MONAI/tutorials/tree/main/bundle/custom_component. Need to update Bundle APP readme on this.
• Bundle [pancreas_ct_dints_segmentation] inference work, training failed. Need to reach out to the creator: Dong.

  File "/monai/transforms/transform.py", line 91, in apply_transform
  return _apply_transform(transform, data, unpack_items)
  File "/monai/transforms/transform.py", line 55, in _apply_transform
  return transform(parameters)
  File "monai/transforms/utility/dictionary.py", line 1040, in __call__
  ret = self._lambd(img=d[key], func=func)
  File "/monai/transforms/utility/array.py", line 832, in __call__
  out = fn(img)
  File "<string>", line 1, in <lambda>
  File "<string>", line 1, in <listcomp>
  TypeError: Cannot interpret 'torch.uint8' as a data type
  RuntimeError: applying transform <monai.transforms.compose.Compose object at 0x7fd1dd3fa1f0>

HI @SachidanandAlle @diazandr3s , I've verified all models in the model zoo. All 3D models for segmentation work well, except DiNTS, reaching Dong on this. For these three: I don't think these should be supported by MONAI Label, it's not typical segmetnation/classification/detection tasks. We need to update Bundle APP readme and keep update which bundle models released in Model-Zoo can be used with MONAI Label. Let me know if this makes sense.

1: [valve_landmarks] 2: [ventricular_short_axis_3label] 3: [mednist_gan]

[tangy5]

Endoscopy app Verfications:

monai version == 1.0.0rc2

Hi @SachidanandAlle @diazandr3s , the endoscopy app is verified on my side, please check these items. If no other items to test, we can click done verification for Endoscopy APP.

• [x] Verified CVAT setup, nuclio container deploy.
• [x] Verified ToolTracking model, inference with CVAT-Nuclio function container, training with MONAI Label. Push MONAI Label fine-tuned model back to Nuclio function container for updating inference model.
• [x] Verified Deid for tool detection, infer with CVAT-Nuclio function container, training with MONAI Label.
• [x] Verified Deepedit for tracking endoscopy images (infer, training).

Note: To improve the Endoscopy APP documentation, mention Troubleshoot issues. To fix some small errors/grammers. Add how tooltracking fine-tuned model can be published to Nuclio function container for updating inference model in CVAT.

[tangy5]

Clients Verification: monai version == 1.0.0rc2 monailabel latest

• [x] 3D Slicer
• [x] 1. Stable Release: 5.0.3
• [x] 2. Preview Release: version 5.1.0, revision 31126 built 2022-09-09
• [x] OHIF
• [x] QuPath
• [x] DSA

[SachidanandAlle]

Lets try to update/add items in the main description also.. so that things can be checked/tracked easily..

[diazandr3s]

Verified Radiology App for all models (infer, train on single/multi-gpu) - done for segmentation, segmentation_spleen, deepgrow 2d+3d+pipeline(infer), deepedit models

Verified vertebra for infer only @diazandr3s please verify once again.. and may be train if that's applicable. Also you need to add one section for this here: https://github.com/Project-MONAI/MONAILabel/tree/main/sample-apps/radiology#model-overview (may be you can combine all models under one section..) and mention about pipeline

Please check https://github.com/Project-MONAI/MONAILabel/pull/996

[tangy5]

QuPath Client verified and done. No issues.

Pathology APP Verification:

I tested Infer/train, with actual Qupath integration and via integration test. There is an issue that loading imagesthat cause run failed. (Confirming with Behrooz)

Problem solved. Openslide's error.

DSA and Pathology app verification check.