Metric Computation

An example Jupyter Notebook is provided which can be run to generate radiomic and dose metrics from the simulation environment data. Once the simulation environment has been deployed, this example demonstrates how to query the CatDB for a cohort of patients, fetch their imaging data from the Orthanc, convert this to a format more suited for our research tools (NIfTI), prepare visualisation of the data and finally compute first-order radiomic features as well as Dose Volume Histograms (DVHs) and some dose metrics. These are saved off to a CSV file, ready for federated analysis.

Deploy Example Docker Image

The easiest way to run this example is to use the Docker Image which contains the example Jupyter Notebook and everything needed to run a Jupyter instance. Modify your Simulation Docker Stack in Portainer by adding the following under services::

  analysis_example:
      image: "auscat/etl:example"
      ports:
      - 8888:8888
      depends_on:
      - etl
      volumes:
      - analysis-data:/data

Also be sure to add the volume definition under volumes: at the end of the stack:

  analysis-data:

Re-deploy the stack.

Running the Example Notebook

Once the Docker image is up and running, you can access the Jupyter instance at: http://[YOUR-HOST-IP]:8888. You will be prompted for a password which has been set to a default of password (Note using this password is only suitable in a simulation environment).

Once logged in, you can navigate to work folder and find the Compute_Metrics.ipynb notebook. Open this Jupyter Notebook and follow the instructions within.

Tip: You can quickly execute the cells within the Jupyter Notebook by pressing Shift+Enter.

Analysing Metrics

The output of this example notebook is a file named metrics.csv which is stored in the working directory. This directory is mapped to the analysis-data volume we defined in the Docker stack above. Using this volume you can proceed with a federated analysis using the metrics computed in this example.

Troubleshooting

Problem: The Jupyter Notebook crashed during conversion of the imaging data.

Solution: This may happen when insufficient memory is available. Ensure you have at least 4GB RAM to run the example. Before re-running the notebook, delete the working folder within the data directory. To do this you may need to open a Terminal by choosing New Launcher then press Terminal and run rm -r work/data/working.