Hi,

I am new to use MONAI and try the tutorial "01 simple_app" by jupyter notebook.

MAP is built successfully and I tried to run but got a error message as following.

the command i used.

Copy a test input file to 'input' folder

!mkdir -p testinput && rm -rf input/* !cp {test_input_path} testinput/ !ls testinput

Launch the app

!monai-deploy run simple_app:latest testinput output

============================================================ the message i got

normal-brain-mri-4.png Checking dependencies... --> Verifying if "docker" is installed...

--> Verifying if "simple_app:latest" is available...

Checking for MAP "simple_app:latest" locally "simple_app:latest" found.

Reading MONAI App Package manifest... Going to initiate execution of operator SobelOperator Executing operator SobelOperator (Process ID: 1, Operator ID: 896fa9c1-0fce-4ab0-a69d-cf86548bfedd) Traceback (most recent call last): File "/opt/monai/app/app.py", line 22, in App(do_run=True) File "/opt/conda/lib/python3.8/site-packages/monai/deploy/core/application.py", line 129, in init self.run(log_level=args.log_level) File "/opt/conda/lib/python3.8/site-packages/monai/deploy/core/application.py", line 429, in run executor_obj.run() File "/opt/conda/lib/python3.8/site-packages/monai/deploy/core/executors/single_process_executor.py", line 125, in run op.compute(op_exec_context.input_context, op_exec_context.output_context, op_exec_context) File "/opt/monai/app/sobel_operator.py", line 21, in compute input_path = next(input_path.glob(".")) # take the first file StopIteration

ERROR: MONAI Application "simple_app:latest" failed.

is there anything wrong and how to solve this, many thanks

Is your feature request related to a problem? Please describe.

Currently, manual work (copy/paste code) is required for a MONAI developer to create a MONAI Deploy App. That is, there is no common API between MONAI and MONAI Deploy (outside of importing MONAI utilities/libraries in a MONAI Deploy App). Due to this, an experienced MONAI developer (primarily focused on developing new algorithms/training paradigms) needs to learn the concepts/API in MONAI Deploy and copy/paste code in the appropriate places in a MONAI Deploy Operator and manually ensure that data transfer concepts (e.g. InferContext) in MONAI Deploy are correctly used.

Analogically, in the case of MONAI Label, an experienced MONAI developer must do much the same: familiarize themselves with the concepts of of MONAI Label (e.g InferTask, TrainTask, etc) and copy/paste their code in the appropriate sections to ensure that the MONAI Label App functions as expected.

Describe the solution you'd like

In addition to researchers "freestyling" training scripts using MONAI, they could also optionally use concepts of MONAI Label and Stream such as tasks (InferTask, TrainTask, etc.) and contexts (InputContext, OutputContext, ExecutionContext) to represent the conceptual task the use aims a certain branch of code to accomplish all the while using data transfer types that are portable across MONAI. The aim is ultimately to employ each of the execution branches for the application the user wants to create; this may still require manual adjustment but overall reduce the amount of manual copy/paste and reduce the learning curve across different MONAI products.

Describe alternatives you've considered Learn each framework in isolation (MONAI, MONAI Label, MONAI Deploy) and copy/paste where appropriate.

This PR is based on the bundle_addtion branch that @ericspod created for automating the parsing of MONAI Bundle and execution of inference, with some notable changes:

• The base Application uses ModelFactory to load model files, e.g. MONAI Bundle TorchScript file, and makes the model network and model file path available through the ExecutionContext. The bundle operator will rely on this mechanism to retrieve the bundle path, and completes the bundle parsing when the operator is executed. The operator does support an optional arg to passing in the bundle path on its constructor so that the parsing can be completed on object creation. The models folder is known to the App at its creation time, but the logic for selecting the specific named model out of potentially multiple models is within the ModelFactory. The App's creation and execution sequence needs to be looked at in future releases so that named model and its path can be retrieved during App creation time so that the bundle path can be reliably passed to the bundle operator on its creation.
• Similar related to the sequence of creation and execution, the bundle operator needs to have its input and output defined on creation so that the App's executor can validate their compatibilities with connected operators. Also, the operator input and output can have IN_MEMORY and DISK storage type, which is information not coming from the bundle, so the application needs to define it on creating the bundle operator instance.
• Added support for parsing metadata from input image and preserving the metadata. This is required so that the inversion in post-processing can work. Also added are special treatment for parsing image object generated from DICOM instances as the order of dims and the names of some metadata are different from what are expected by the transforms.
• Although the bundle operator will typically be used in between other operators, e.g. DICOM image converter and seg image to DICOM writer operators, a user might use the bundle operator as the first or leaf operator. For such cases, the app executor does have restrictions that the storage type needs to be DISK, so the input or output needs to be files. In order not to have many file formats to support in this initial release of the bundle operator, only Python pickle file is supported. Pickle file does have some security concerns, though content is only used as pure data object, validated after loading, thus mitigating the risk.
• The bundle operator can be used in an application that has multiple models, by specifying the model name (the ModelFactory has a specification on model folder structure when multiple models are used).
• The bundle operator and the example app have been tested with the Spleen_CT_Segmentation bundle from MONAI Model Zoo.

I am recreating this pull request since I completely borked #298 due to the (extremely frustrating) DCO sign off requirement and subsequent attempts to fix it that made it worse :persevere:

This PR introduces the highdicom library to handle the creation of DICOM segmentations. This should lead to more maintainable code by using the well-tested highdicom package for the complex task of segmentation creation, and should make adding new features to the segmentation writer operator considerably easier by leveraging the prior work in that library. Note that the complexity of the ad-hoc code in the dicom segmentation operator file is reduced significantly.

Unlike the original version of #298 I have now removed the highdicom class highdicom.seg.SegmentDescription from the public API of the monai deploy class. However I have left the pydicom.sr.coding.Code class in the public API as we discussed in the last meeting.

I have not yet changed the examples to use the new version of the operator, as this requires merging with some recent changes due to the bundle that I still need to get my head around. This will need to be addressed before merging so that the examples run. But I would like to get agreement on the API of the class before fixing the examples.

Update: All example apps that make use of the Segmentation Writer have been updated in this PR, and all used the appropriate codes. Also the monai core is pinned at v0.9.0 because the latest version 0.9.1 is not compatible with the apps (not related to the new Seg Writer itself)

Addresses #195 (partially)

Introduces the highdicom library to handle the creation of DICOM segmentations. This should lead to more maintainable code by using the well-tested highdicom package for the complex task of segmentation creation, and should make adding new features to the segmentation writer operator considerably easier by leveraging the prior work in that library. Note that the complexity of the ad-hoc code in the dicom segmentation operator file as reduced signficantly.

I am considering this a largely complete "minimum viable" implementation and would appreciate feedback at this stage about whether we should go forward with this and what may need to change. Note that this is a backwards incompatible change to the signature of the DICOMSegmentationWriterOperator's __init__ method. It is now necessary to pass a list of highdicom.seg.SegmentDescription objects to the constructor to describe each segment in the created segmentation. I'd argue that the existing implementation, which hard codes in some default values, is likely to create dicom segs that are semantically incorrect in many cases (e.g. hard coding the segmented category as organ when it may be an abnormality such as a tumor or nodule). I am assuming that this backwards incompatibility is acceptable at this stage since the docstring states that the interface may change. Would appreciate thoughts on this.

I have attempted to update the two examples that use the seg writer: spleen and liver tumor segmetentations. I have managed to test the spleen example and visualize the results, and it's working nicely. I have not managed to test the liver tumor example. I may be being stupid but I can't find where to download the relevant files to actually run it.

Another thing I'm a bit confused about is where and how to add highdicom as an (optional) requirement. There are lots of requirements files and I wasn't totally sure how it all fits together. For now I just added highdicom to requirements_dev.txt. Please let me know whether this is correct.

My apologies that this took much longer than I anticipated to make happen (I largely blame this monster). Now I've figured out how this stuff all fits together, adding further operators (such as SR, GSPS) using the capabilities in highdicom should become much easier if we choose to merge this.

Tagging highdicom author and co-maintainer @hackermd for visibility.

Describe the bug We are working on deployment of spleen segmentation using monai deploy, we are using pretrained models from ngc for spleen segmentation from (https://catalog.ngc.nvidia.com/orgs/nvidia/teams/med/models/clara_pt_spleen_ct_segmentation) link.

Similarly we are using the app.py from examples in deploy-app-sdk(https://github.com/Project-MONAI/monai-deploy-app-sdk/tree/main/examples/apps/ai_spleen_seg_app). when we run python app.py -i ./dcm -o ./output -m ./model.pt .

We get the following error

raise ItemNotExistsError("A predictor of the model is not set.") monai.deploy.exceptions.ItemNotExistsError: A predictor of the model is not set.

We tried our own trained model and still got the same error.

Steps/Code to reproduce bug

1. Used app.py from examples in monai-deploy-app-sdk(https://github.com/Project-MONAI/monai-deploy-app-sdk/tree/main/examples/apps/ai_spleen_seg_app)
2. Run the python app.py -i ./dcm -o ./output -m ./model.pt
   1. Input images from ai_spleen_seg_data which is provided in the MONAI Deploy Segmentation App documentation
   2. Used pretrained model from ngc for spleen segmentation from (https://catalog.ngc.nvidia.com/orgs/nvidia/teams/med/models/clara_pt_spleen_ct_segmentation) link.

Expected behavior We expect the pretrained model from the link , also results in the output without any issues.

Environment details (please complete the following information)

• OS/Platform: linux 4.15.0-136-generic
• Python Version: Python 3.8.10 | packaged by conda-forge
• Method of MONAI Deploy App SDK install: [pip, conda, Docker, or from source] : pip install monai-deploy-app-sdk
• SDK Version: 0.2.0

Additional context This works fine for the spleen model downloaded from google drive link (https://drive.google.com/uc?id=1uTQsm8omwimBcp_kRXlduWBP2M6cspr1) which was in the tutorial MONAI Deploy Segmentation App documentation

Is there any specific thing we need to do while saving and using the model with deploy app sdk.

Any suggestion and help is appreciated. Thank you

User Story As an application developer I would like to be able to select a subset of DICOM Series given one or more DICOM Studies. My selection criteria are based on conditions imposed on DICOM attributes embedded in the SOP instances that belong to one or more series

Background Often an app developer needs to do some work to identify what type of medical images are applicable given an AI model. As data flows in from sources such as modality or PACS, it becomes challenging to automatically select what data needs to be provided so that an AI model can ingest to produce correct result. As of v0.1.0, this operator always returns the first series given a Study

Success Criteria

• A way to specify selection criteria using key value pairs of DICOM Attributes
• A way to parse through DICOM datasets and identify the right study --> series --> sop instance sets given the criteria
• An example application that showcases such series selection using this operator
• Updated API documentation for the operator
• Unit Test cases that exercise multiple scenarios for this Operator

Describe the bug

Steps/Code to reproduce bug

• pip install monai v0.9.1
• run the existing example apps, Spleen, LiverTumor, UNETR etc. Expected behavior Ideally monai v0.9.1 should not have broken existing consumers, especially when it is supposed to be largely backward compatible.

Given that monai v0.9.1 has been released and published, the App SDK needs to declare incompatibility with this version until the App SDK implementation is updated and tested working with monai 0.9.1.

Environment details (please complete the following information)

• OS/Platform: Ubuntu 20.04 LTS, GPU Nvidia GV100 with 32 GB mem
• Python Version: 3.8
• Method of MONAI Deploy App SDK install: [pip install from source]
• SDK Version: main branch

Additional context See attached log files from running the apps (to be attached), with a brief snippet shown below:

Traceback (most recent call last):
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/transforms/transform.py", line 90, in apply_transform
    return _apply_transform(transform, data, unpack_items)
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/transforms/transform.py", line 54, in _apply_transform
    return transform(parameters)
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/transforms/post/dictionary.py", line 642, in __call__
    transform_info = d[InvertibleTransform.trace_key(orig_key)]
KeyError: 'image_transforms'

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/home/mqin/src/monai-app-sdk/.venv/bin/monai-deploy", line 33, in <module>
    sys.exit(load_entry_point('monai-deploy-app-sdk', 'console_scripts', 'monai-deploy')())
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/deploy/cli/main.py", line 116, in main
    execute_exec_command(args)
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/deploy/cli/exec_command.py", line 76, in execute_exec_command
    runpy.run_path(app_path, run_name="__main__")
  File "/usr/lib/python3.8/runpy.py", line 265, in run_path
    return _run_module_code(code, init_globals, run_name,
  File "/usr/lib/python3.8/runpy.py", line 97, in _run_module_code
    _run_code(code, mod_globals, init_globals,
  File "/usr/lib/python3.8/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/home/mqin/src/monai-app-sdk/examples/apps/ai_livertumor_seg_app/app.py", line 90, in <module>
    app_instance.run()
  File "/home/mqin/src/monai-app-sdk/examples/apps/ai_livertumor_seg_app/app.py", line 37, in run
    super().run(*args, **kwargs)
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/deploy/core/application.py", line 429, in run
    executor_obj.run()
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/deploy/core/executors/single_process_executor.py", line 125, in run
    op.compute(op_exec_context.input_context, op_exec_context.output_context, op_exec_context)
  File "/home/mqin/src/monai-app-sdk/examples/apps/ai_livertumor_seg_app/livertumor_seg_operator.py", line 105, in compute
    infer_operator.compute(op_input, op_output, context)
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/deploy/operators/monai_seg_inference_operator.py", line 226, in compute
    d = [post_transforms(i) for i in decollate_batch(d)]
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/deploy/operators/monai_seg_inference_operator.py", line 226, in <listcomp>
    d = [post_transforms(i) for i in decollate_batch(d)]
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/transforms/compose.py", line 173, in __call__
    input_ = apply_transform(_transform, input_, self.map_items, self.unpack_items, self.log_stats)
  File "/home/mqin/src/monai-app-sdk/.venv/lib/python3.8/site-packages/monai/transforms/transform.py", line 114, in apply_transform
    raise RuntimeError(f"applying transform {transform}") from e
RuntimeError: applying transform <monai.transforms.post.dictionary.Invertd object at 0x7f5e3e4ac4f0>

https://github.com/Project-MONAI/monai-deploy-app-sdk/blob/8ec336fffffbaf54cbe3464540f4e36763d92360/monai/deploy/operators/monai_seg_inference_operator.py#L211

Applications that use the Dataloader in MONAI Seg Inference Operator throw errors when trying to allocate shared memory in Argo pipelines.

Need to modify the num of workers argument to not use shared memory by default.

Describe the bug Codes used to describe segments in DICOM SEG writer should be improved.

Steps/Code to reproduce bug See https://github.com/Project-MONAI/monai-deploy-app-sdk/blob/294fb6e185921a621ad28754d5b97999485e803e/monai/deploy/operators/dicom_seg_writer_operator.py#L440-L463.

Expected behavior

1. SegmentationCategoryType should refer to the code corresponding to the specific organ. Currently, there is a general code for "Organ", with the specifics about which organ is segmented listed in the SegmentLabel. This is not the intended use of this object.
2. Codes should be using SNOMED-CT SCT codes, currently used SRT codes have been deprecated in DICOM.

Environment details (please complete the following information)

• SDK Version: current as of submitting

Additional context

• Relevant commit in dcmqi: https://github.com/QIICR/dcmqi/commit/0cf28290be3ab15afc818b3d1d032f70a8c782a0
• why don't you just use highdicom to write DICOM SEG? https://github.com/herrmannlab/highdicom

Hello everyone, After getting the latest main branch and upgrading monai deploy. Running the 03_segmentation_viz_app gives the following error

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
/tmp/ipykernel_510135/3062323273.py in <module>
     27 from monai.deploy.operators.dicom_series_selector_operator import DICOMSeriesSelectorOperator
     28 from monai.deploy.operators.dicom_series_to_volume_operator import DICOMSeriesToVolumeOperator
---> 29 from monai.deploy.operators.clara_viz_operator import ClaraVizOperator

ModuleNotFoundError: No module named 'monai.deploy.operators.clara_viz_operator'

I checked that in my folder operators there is the file clara_viz_operator.py. In addition I checked init.py and there exist the required line

 29 from .clara_viz_operator import ClaraVizOperator

I donot understand what is going on? Any comments @AndreasHeumann

I am trying to run this example: https://docs.monai.io/projects/monai-deploy-app-sdk/en/latest/notebooks/tutorials/03_segmentation_app.html

but I am getting this error importing the libraries: import logging from os import path

from numpy import uint8

import monai.deploy.core as md from monai.deploy.core import ExecutionContext, Image, InputContext, IOType, Operator, OutputContext from monai.deploy.operators.monai_seg_inference_operator import InMemImageReader, MonaiSegInferenceOperator from monai.transforms import ( Activationsd, AsDiscreted, Compose, EnsureChannelFirstd, EnsureTyped, Invertd, LoadImaged, Orientationd, SaveImaged, ScaleIntensityRanged, Spacingd, )

Required for setting SegmentDescription attributes. Direct import as this is not part of App SDK package.

from pydicom.sr.codedict import codes

from monai.deploy.core import Application, resource from monai.deploy.operators.dicom_data_loader_operator import DICOMDataLoaderOperator from monai.deploy.operators.dicom_seg_writer_operator import DICOMSegmentationWriterOperator, SegmentDescription from monai.deploy.operators.dicom_series_selector_operator import DICOMSeriesSelectorOperator from monai.deploy.operators.dicom_series_to_volume_operator import DICOMSeriesToVolumeOperator

ImportError Traceback (most recent call last) /tmp/ipykernel_1969259/585954608.py in 26 from monai.deploy.core import Application, resource 27 from monai.deploy.operators.dicom_data_loader_operator import DICOMDataLoaderOperator ---> 28 from monai.deploy.operators.dicom_seg_writer_operator import DICOMSegmentationWriterOperator, SegmentDescription 29 from monai.deploy.operators.dicom_series_selector_operator import DICOMSeriesSelectorOperator 30 from monai.deploy.operators.dicom_series_to_volume_operator import DICOMSeriesToVolumeOperator

ImportError: cannot import name 'SegmentDescription' from 'monai.deploy.operators.dicom_seg_writer_operator' (/home/sonosa/anaconda3/envs/monai/lib/python3.7/site-packages/monai/deploy/operators/dicom_seg_writer_operator.py)

Describe the bug I am following the example of https://docs.monai.io/projects/monai-deploy-app-sdk/en/0.2.1/getting_started/tutorials/02_mednist_app.html

I was able to run everything but the part of the docker: monai-deploy run mednist_app:latest input output

I am getting this error:

(monai) sonosa@sonosa-MS-7B17:~/2022/ProjectsAI/monai$ monai-deploy run mednist_app:latest input output_docker_gpu /home/sonosa/anaconda3/envs/monai/lib/python3.7/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: /home/sonosa/anaconda3/envs/monai/lib/python3.7/site-packages/torchvision/image.so: undefined symbol: _ZNK3c1010TensorImpl36is_contiguous_nondefault_policy_implENS_12MemoryFormatE warn(f"Failed to load image Python extension: {e}") Checking dependencies... --> Verifying if "docker" is installed...

--> Verifying if "mednist_app:latest" is available...

Checking for MAP "mednist_app:latest" locally "mednist_app:latest" found.

Reading MONAI App Package manifest... --> Verifying if "nvidia-docker" is installed...

/opt/conda/lib/python3.8/site-packages/scipy/init.py:138: UserWarning: A NumPy version >=1.16.5 and <1.23.0 is required for this version of SciPy (detected version 1.23.5) warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion} is required for this version of " Going to initiate execution of operator LoadPILOperator Executing operator LoadPILOperator (Process ID: 1, Operator ID: a37656f0-65c6-46ca-8fdf-e8477bab45d2) Done performing execution of operator LoadPILOperator

Going to initiate execution of operator MedNISTClassifierOperator Executing operator MedNISTClassifierOperator (Process ID: 1, Operator ID: c9f285dd-3b8c-4ef9-b700-cd6ac49186a0) /root/.local/lib/python3.8/site-packages/monai/utils/deprecate_utils.py:107: FutureWarning: <class 'monai.transforms.utility.array.AddChannel'>: Class AddChannel has been deprecated since version 0.8. please use MetaTensor data type and monai.transforms.EnsureChannelFirst instead. warn_deprecated(obj, msg, warning_category) /root/.local/lib/python3.8/site-packages/monai/utils/type_conversion.py:134: UserWarning: The given NumPy array is not writeable, and PyTorch does not support non-writeable tensors. This means you can write to the underlying (supposedly non-writeable) NumPy array using the tensor. You may want to copy the array to protect its data or make it writeable before converting it to a tensor. This type of warning will be suppressed for the rest of this program. (Triggered internally at /opt/pytorch/pytorch/torch/csrc/utils/tensor_numpy.cpp:175.) tensor = torch.as_tensor(tensor, kwargs) device found : cuda terminate called after throwing an instance of 'c10::Error' what(): isTuple()INTERNAL ASSERT FAILED at "/opt/pytorch/pytorch/aten/src/ATen/core/ivalue_inl.h":1397, please report a bug to PyTorch. Expected Tuple but got String Exception raised from toTuple at /opt/pytorch/pytorch/aten/src/ATen/core/ivalue_inl.h:1397 (most recent call first): frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits, std::allocator >) + 0x6c (0x7f27560b224c in /opt/conda/lib/python3.8/site-packages/torch/lib/libc10.so) frame Project-MONAI/MONAI#1: c10::detail::torchCheckFail(char const, char const, unsigned int, std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&) + 0xfa (0x7f275607da66 in /opt/conda/lib/python3.8/site-packages/torch/lib/libc10.so) frame Project-MONAI/MONAI#2: c10::detail::torchInternalAssertFail(char const*, char const*, unsigned int, char const*, std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&) + 0x53 (0x7f27560b0233 in /opt/conda/lib/python3.8/site-packages/torch/lib/libc10.so) frame Project-MONAI/MONAI#3: + 0x4224e29 (0x7f27a23b4e29 in /opt/conda/lib/python3.8/site-packages/torch/lib/libtorch_cpu.so) frame Project-MONAI/MONAI#4: + 0x42253e9 (0x7f27a23b53e9 in /opt/conda/lib/python3.8/site-packages/torch/lib/libtorch_cpu.so) frame Project-MONAI/MONAI#5: torch::jit::SourceRange::highlight(std::ostream&) const + 0x48 (0x7f279f3d5c58 in /opt/conda/lib/python3.8/site-packages/torch/lib/libtorch_cpu.so) frame Project-MONAI/MONAI#6: torch::jit::ErrorReport::what() const + 0x2c3 (0x7f279f3baac3 in /opt/conda/lib/python3.8/site-packages/torch/lib/libtorch_cpu.so) frame Project-MONAI/MONAI#7: + 0x9ea44f (0x7f27a873344f in /opt/conda/lib/python3.8/site-packages/torch/lib/libtorch_python.so) frame Project-MONAI/MONAI#8: + 0x9fa12d (0x7f27a874312d in /opt/conda/lib/python3.8/site-packages/torch/lib/libtorch_python.so) frame Project-MONAI/MONAI#45: __libc_start_main + 0xf3 (0x7f27e722d083 in /usr/lib/x86_64-linux-gnu/libc.so.6)

ERROR: MONAI Application "mednist_app:latest" failed.

=================================================================

Please any help could be useful, Just in case I tested the nvidia-docker and it is working very well

Add operators to write out GSPS DICOM objects for:

• Object detection (bounding box)
• Segmentations (overlay boundaries)

We can leverage highdicom to do this

Describe the bug An prostate segmentation application is built with App SDK 0.5 and a model trained with MONAI core v1. It takes in a MR T2 series, creates the segmentation image, which then get written as DICOM Seg as well as transformed into a STL image. The former looks good, while latter looks like a squished donut.

Steps/Code to reproduce bug Application code is supper simple since it uses the built-in MONAI bundle inference operator as well as others, though the model itself is proprietary, not to be shared publicly.

Expected behavior The STL generated from the seg image (numpy + metadata) should represent closely the seg image

Environment details (please complete the following information)

• OS/Platform: Ubuntu 20.04 LTS
• Python Version: 3.8
• Method of MONAI Deploy App SDK install: [pip, and from source]
• SDK Version: 0.5

Additional context The STL operator has been tested with the liver and spleen application and generated the correctly mesh image. Noted that those seg images have hundreds of slices, while the prostate MR has only around 20 slices, though, it is expected the 3D spacings should be correct irrespective of number of slices.

Is your feature request related to a problem? Please describe. With the use of logging library, application execution logs can be gathered at various logging level per configuration. However, with the maturing of the App SDK, especially when venturing into clinical deployment (IRB'ed), there are stronger needs to provide additional insights into the application execution, e.g. activity audits, performance metrics both in terms of latency and ideally result accuracy.

This is important when AI applications are deployed and used in imaging workflow, as the IT/PACS admins always need to have insights into the latency, resource, and clinical performance of the applications, so as to better manage the critical workflow stage for operational, clinical, as well as regulatory requirements.

Describe the solution you'd like A logging framework that will support pre-defined as well as custom categories of metrics collection, while making the aggregation and delivery of the metrics dynamic and configurable, e.g. records gather could be delivered to the stats gathering agent on the host system that runs the MAP.

Describe alternatives you've considered Native logging library with logging level combined with defined string literal indicating the categories of information, e.g. default being execution log, "PERF" for elapsed time for blocks of execution, etc.

Additional context

Is your feature request related to a problem? Please describe. Often we would like to know how the model is performing on the data over a period of time and maintain overall statistics of the model performance. These statistics can be further granularized to specific centers and demographics. If several models are deployed we can use this performance to track the performance of each separate model.

Describe the solution you'd like A web-based dashboard where the model performance can be displayed. Something like Tensorboard.

Describe alternatives you've considered None so far.

Additional context These statistics can be further used for design model updating rules in a continuous fashion.