tamara-glazer/med-image-classifier

Detect and classify potentially cancerous legions in mammogram images

Classifying abnormal masses in mammograms

Tammy Glazer, Parth Khare, Cecile Murray

Downloading data

1. Add desired images to the cart and download. get-mass-case-ids.sh will help create a comma-separated list of training and testing cases using the metadata csvs.

2. Use NBIA data retriever to convert images from .tcia to .dcm (longest step)

3. Run sh ingest-data.sh followed by the full file path to the images to pull .dcm files out of nested directory and sensibly rename.

Setting up and using the virtual environment

If you use conda, you might have to conda deactivate before these steps.

First, create the virtual environment:

python -mvenv venv_name

Then activate it:

source ./venv_name/bin/activate

Then install required packages based on the list in requirements.txt:

pip3 install -r requirements.txt

You can install additional packages as usual. To add them to the list of required packages, you can run:

pip3 freeze > requirements.txt