The DL mammography project

This repository stores the code used for the article "Deep learning predicts interval and screen-detected cancer from screening mammograms: a case-case-control study".

Prerequisites

• Linux Kernel 4.4+
• Python 3.6+, with recent versions of the following python packages:
  • cv2
  • fire
  • matplotlib
  • numpy
  • pandas
  • pydicom
  • scipy
  • seaborn
  • pytorch

Running

The pipeline consists of three steps, which are separated into three folders.

./s1-data_preprocessing
./s2-cnn_model
./s3-GAN

Inside each folder, there is a Python script which serves as the entry point. These are:

./s1-data_preprocessing/preprocess.py
./s2-cnn_model/run.py
./s3-GAN/run.py

Running the entry point script without any command line arguments will prompt a list of required/optional arguments. These script assumes that raw files were put into the relative path of ./raw_data and their outputs are put into the relative path of ./out_data, both are relative to the entry point's directory.