Web-based DICOM viewer with labeling tool

This repository contains a simple web-based DICOM viewer built using Streamlit. The viewer also has a tool to annotate which slices of a specific series of dicom files have anomalies. Annotations are in .JSON format with labels "Anomaly" and "Slices" for each series. This tool was meant to be a weekend project for me to learn how to work with Streamlit, Docker and Heroku, but it took a little more than that to figure out how to set up everything and to make it look good.

Demo hosted on Heroku Render (Now that Heroku has canceled its free-tier)

The tool only deals with zip files that have one or more folders, normally represented as series, with dicom files inside. A sample of an acceptable zip file can be checked inside the sample folder. Zip files can be uploaded via public shared URLs from Google Drive or using the file upload widget.

As the demo is hosted on Heroku and their free tier dyno has limited memory resources, uploaded zip files are limited to 100MB. However, this restriction can be adjusted by code when running the web viewer locally.

Setup

Docker

Clone the repository and set the current directory:

git clone https://github.com/angelomenezes/dicom-labeling-tool.git
cd dicom-labeling-tool/

For running a local docker container, change the line in the Docker file from:

CMD streamlit run DICOM.py --server.port $PORT

to

CMD streamlit run DICOM.py --server.port 8501

Then, you can build and run successfully the container with:

docker build ./ --tag webapp:v1
docker container run -p 8501:8501 webapp:v1

To finish, open the browser at http://localhost:8501/

Conda

Make sure you have Anaconda installed since it is the easiest way to setup GDCM on Python3 which is a requirement for the pydicom library.

(Optional) Create a conda environment for installing and running the app:

conda create --name DICOM_env python=3.6.10 -y
conda activate DICOM_env

Clone the repository and set the current directory:

git clone https://github.com/angelomenezes/dicom-labeling-tool.git
cd dicom-labeling-tool/webapp/

conda install -c conda-forge -y gdcm
pip install -r requirements.txt

streamlit run DICOM.py

To finish, open the browser at http://localhost:8501/

License

MIT

Contact

Any comments, suggestions or contributions are welcome. You can contact me at [email protected].