Detection and effect of artifacts in breast mammography

This repository contains the code associated with the paper "Radio-opaque artefacts in digital mammography: Automatic detection and analysis of downstream effects".

It contains the following files:

• labelling_tools contains the notebook with the lightweight artifact labelling tool
• artifact_detector_model.py contains the model definition for the multi-label artifact detector
• artifact_train.py contains the code to train the detector
• artifact_evaluation.ipynb contains the evaluation code/plotting for the detector
• downstream_model.py contains the model definition for the downstream evaluation tasks (lesion detection and density prediction)
• cancer_train.py to train the screening outcome / lesion detection prediction model
• density_train.py to train the density classification model
• dataset.py defines dataset classes and pytorch lightining data modules for all training tasks.

Artifact datasets

The manually labelled artifact dataset file can be found in labelling_tools/manual_annotations_new.csv. The model predictions from the artefact detector for all images in EMBED can be found in predicted_all_embed.csv

Requirements

All required pip depencies needed to run code in this project are listed in requirements.txt

Train the artifact detector

Simply run python train_detector.py to train your own artefact detector.

Train the downstream model (e.g. density)

Simply run python density_train.py to train a density classification model.

To assess the model per artifact you can then run inference: python density_inference.py and analyse outputs with density_evaluate_markers.ipynb