dicom-flask-uploader

Python Web Server in Flask that allows users to upload & browse DICOM images. Under the hood mudicom and gdcm are used to get key/value info from the DICOM files. The pillow library is used to make the thumbnail.

See an example running at:

http://dicom-flask-uploader.herokuapp.com/

How to run

Using conda + pip

To Deploy, install the dependencies

conda create -p env/
source activate env/
# You need to install gdcm
conda config --add channels conda-forge
conda install --file requirements/conda.txt
pip install -e .

Then run

export FLASK_APP=dicom_flask_uploader.runapp
flask initdb
flask importdb
flask run

To Run Tests

source activate env/
pip install -r requirements/development.txt
py.test

Design choices

• Using simple sqlite db at the moment for quick prototyping, but could be changed for any other type of relational db that flask-sqlalchemy supports.
  • Data element key/value sizes picked arbitrarily. Not sure what max limits are for DICOM
  • Might actually be better to use nosql db for this, since key/values can be different between DICOMs
• Use gdcm so compressed DICOM images can be handled
  • There was a conda binary already available, so was easy to use that. Saw there were some Dockerfile images available as well
• Storing thumbnails/dcm on /tmp/ so I don't get charged for uploading/storing files
  • For persistent storage on Heroku, could've also stored them in the db but that is generally considered bad practice
• Using Jinja2 templating with flask for quick prototyping
• A few tests to show system works, no reason to test a prototype super thoroughly

Further Ideas

• Use that lovely flask-resty library to create REST API
• Add authentication
• Store static files in a persistent manner using flask-annex
• Seems like some thumbnails are not always properly converted, probably issue in mudicom or gdcm
• Add better error checking (for uploading incorrect image type for instance)
• Validate DICOM files and show info to user
• Query DICOM key/value