Annotate and Adapt Medical Imaging Models with MONAI
Annotate and Adapt Medical Imaging Models with MONAI (Latest Version)

Step #2: Setting Up the QuPath viewer

There are several viewers that can be integrated into the Pathology app. We created a plugin in QuPath to show how nuclei segmentation in whole slide images (WSIs) can be performed using MONAI Label + QuPath.

QuPath is an open software for bioimage analysis. it can be used as an efficient tool for the annotation and visualization of whole slide images.

The first step is to go to the Desktop and open QuPath.

pathology-019.png

For this, open the QuPath folder located on the Desktop:

pathology-005.png

Then, open the bin folder:

pathology-002.png

Finally, start a terminal by right-clicking on the folder:

pathology-003.png

and run QuPath by typing in the terminal ./QuPath and then “Enter”:

pathology-004.png

Once the QuPath is up and running, the next step is to drag the jar file (located on the Desktop) on top of the running QuPath application window (black screen area) to install the extension. See the following images:

pathology-001.png
pathology-020.png
pathology-006.png

If this message appears, you could overwrite it, by clicking “OK”:

pathology-021.png

On the menu bar, navigate and click MONAI Label. Click Next Sample/Patch to load data specified in the start_server command. If the QuPath starts from the beginning, the button will automatically load the next whole slide image.

pathology-007.png

Click on Apply for default image type (Brightfield H&E)

pathology-022.png

And then OK

pathology-023.png

Users can select the blue rectangle to pick an ROI.

pathology-024.png
pathology-025.png

Then select the segmentation_nuclei model in the pop-up interface, and click OK to run inference.

pathology-026.png
pathology-027.png
Note

MONAI Label server will run auto segmentation with the selected patch data and return the segmentation with polygons.


Users can edit the predicted nuclei segmentation to retrain the model. Once the correction is done, users can navigate to the MONAI Label plugin and click Submit Label to save the ground truth.

pathology-028.png

Then click Yes to save the ROI label:

pathology-029.png

MONAI Label server provides Nuclic models to perform instance segmentation with a single click.

pathology-015.png
pathology-016.png

Once users finish annotations on several patch images, MONAI Label provides a training option to fine-tune the model:

pathology-017.png

In the training option pop-up interface, select segmentation_nuclei, and select training customized training options. Uncheck pretrained to start training the model from scratch. Finally, click OK to start the model training in the MONAI Label server:

pathology-018.png

Repeat fetching and segmenting nuclei until a satisfied model is achieved or entire datasets/WSIs are annotated.

Now that you have a working MONAI Label implementation, work through our Quick Start Guide and/or the resources available on the Next Steps page.

Note

If you are done with the MONAI Label lab, just close the browser window/tab


© Copyright 2022-2023, NVIDIA. Last updated on Apr 26, 2023.