cnn-bcdr

Source code for Representation learning for mammography mass lesion classification with convolutional neural networks (pdf).

Dependencies

This code is written in python. To use it you will need:

• Python 2.7
• Pylearn2 (tested on pylearn2@30437ee)
• Scipy
• [Shapely] (https://github.com/Toblerity/Shapely)

Usage

Data

With this paper we released the Breast Cancer Digital Repository F03 (BCDR-F03) dataset. You can get a copy from http://bcdr.inegi.up.pt/ (mirror at: https://bcdr.ceta-ciemat.es). Uncompress it under the data folder.

Preprocessing

• Create hdf5 dataset:

  python make_dataset.py config.yaml
  

• Build preprocessed version (GCN + LCN):

  python preprocessing.py config.yaml
  

Training

The hyperparameters to train the network are in the config.yaml file. Train the model:

python train_cnn.py config.yaml

Evaluation

Evaluate trained model:

python eval.py config.yaml