[VDR ]add ohif brief overview doc in the readme #84
Conversation
|
|
||
| To begin using the NVIDIA MONAI Cloud APIs, you can sign up for early access by filling out this survey: [NVIDIA MONAI Cloud API Early Access](https://developer.nvidia.com/nvidia-monai-cloud-api-early-access-program/join). | ||
|
|
||
| ## Getting Started with OHIF Plugin |
There was a problem hiding this comment.
If we folllow the style @zephyrie 's integrated_doc branch, I would suggest to move the contents of OHIF and DICOMweb to two markdown files.
There was a problem hiding this comment.
Yeah, we should break this down so people can find the relevant section and we should not have this in the README but in the docs section for How-To guides.
"DICOMWeb Setup" and "Using OHIF and MONAI Cloud API Plugin", we can also move the instructions from the OHIF plugin setup directory to another docs page "Setting up OHIF and MONAI Cloud API Plugin"
There was a problem hiding this comment.
Also, can we update the plugin name / any other displayed name to "NVIDIA MONAI Cloud API"? We should stop using MONAI Service for any naming moving forward.
|
|
||
| ## Getting Started with OHIF Plugin | ||
|
|
||
| MONAI Service provide ready-to-use playgroundw with API server and applications for users. |
There was a problem hiding this comment.
The whole sentence can be
The MONAI Service offers a ready-to-use environment featuring an API server and applications for users. If you wish to construct and customize the OHIF viewer along with the MONAI Service plugin, please consult the [setup page](https://github.com/NVIDIA/monai-cloud-api/tree/main/plugins/ohif).
| If you are going to build and custimize OHIF viewer and MONAI Service plugin. Please refer to the [setup page](https://github.com/NVIDIA/monai-cloud-api/tree/main/plugins/ohif) | ||
|
|
||
| ### Dicomweb Overview | ||
| Dataset storage can be variant, in the open-sourced setup. We recommend Orthanc for hosting dicom format medical images. |
There was a problem hiding this comment.
The storage of datasets can vary; in our open-source setup, we recommend using Orthanc for hosting DICOM format medical images.
| sudo apt-get install orthanc orthanc-dicomweb -y | ||
|
|
||
| # Install Plastimatch | ||
| sudo apt-get install plastimatch -y |
There was a problem hiding this comment.
Can these two be merged into one-line?
sudo apt-get install orthanc orthanc-dicomweb plastimatch -y
There was a problem hiding this comment.
As far as I can tell, we don't need plastimatch for orthanc, but for the conversion of nifti to dicom? We can leave it as a separate code block, but we should give me information on why we need it, give an example of using it to convert, and also specify the installation is optional and only required for the conversion of files.
| For customized setup for Orthanc, | ||
|
|
||
| **Steps:** | ||
|
|
There was a problem hiding this comment.
Add
1. Install Orthanc and DICOMweb plugin
| ``` | ||
|
|
||
| ``` | ||
| dicom_web_endpoint = ... |
There was a problem hiding this comment.
Where will these variables be consumed? MONAI cloud api notebooks?
|
|
||
|
|
||
| #### 1. **Study List** | ||
| Open OHIF viewer, the index page is a study list. The subjects listed are prepared in dicom-web storage. |
There was a problem hiding this comment.
Incomplete sentence:
Open OHIF viewer, the index page is a study list
|
|
||
|
|
||
|
|
||
| #### 4. **Run Inferencing Using Selected Method** |
There was a problem hiding this comment.
After this MR is merged, we can remove the duplicated contents in the notebook: https://github.com/NVIDIA/monai-cloud-api/blob/main_ea4/notebooks/Annotation%20and%20Continuous%20Learning%20Overview.ipynb
There was a problem hiding this comment.
Yeah, we should move as much as possible that's duplicate or unnecessary for the actual flow of the notebook to docs.
| **Class Prompts**: | ||
|
|
||
| The class prompts is used when inference and annotating specific anatomies, VISTA model supprts more | ||
| than 100 organs, bones, muscles, etc. Users can selecte one or multiple targets for AI inference. |
| Use mouse and click to place the control points within medical images. | ||
| In OHIF MONAI Service plugin, user can use point prompts by clicking the | ||
| **Point Prompts**: | ||
| You can see availabel organs in the panel, select one anatomy for point |
|
|
||
|  | ||
|
|
||
| Next, notify the server that an image has been annotated. This step is crucial for continuous learning. The system will take note of the new annotations and after the indicated number of annotated images it will use them to improve the model over time. |
There was a problem hiding this comment.
continuous learning -> continual learning
| This step ensures that your annotations are as accurate as possible. | ||
|
|
||
| **Steps** | ||
| 1. Click the Segmentation button. |
|
|
||
| **Steps** | ||
| 1. Click the Segmentation button. | ||
| 2. Select a class of segmentation that needs to be updated. |
| - Refine boundaries | ||
| - Add or remove regions | ||
|
|
||
| This step ensures that your annotations are as accurate as possible. |
There was a problem hiding this comment.
This step is crucial to ensure the highest possible accuracy of your annotations.
|
|
||
| **Class Prompts**: | ||
|
|
||
| The class prompts is used when inference and annotating specific anatomies, VISTA model supprts more |
There was a problem hiding this comment.
"The class prompt is" or "The class prompts are"
|
|
||
|  | ||
|
|
||
| User can get the real-time prompt response for the class edits. |
There was a problem hiding this comment.
Would be good to add a small summary or conclusion after this line.
|
|
||
|  | ||
|
|
||
| The associated API call run when you click the `Notify Server` button is below: |
There was a problem hiding this comment.
Remove, we don't need to inform people during this guide what the corresponding API call is. We should define a Viewer Integration guide that talks about the dataset API calls and how they should be used.
| @@ -14,6 +15,191 @@ The NVIDIA MONAI Cloud APIs offer a comprehensive suite of tools designed to enh | |||
| - A robust foundation model, trained on 117 classes encompassing various organs and tumors. | |||
| - The model's training and validation involved CT images from over 4,000 patients, ensuring its high accuracy and reliability for medical image analysis. | |||
|
|
|||
| ### How to Get Started | |||
| 4. Dicon-web Setup and Dataset Configuration: | |||
|
Closing as the doc is available elsewhere. |
Fix #65 , #66