haarburger / multi-scale-curriculum

Code for Paper: Multi Scale Curriculum CNN for Context-Aware Breast MRI Malignancy Classification

Multi Scale Curriculum CNN for Context-Aware Breast MRI Malignancy Classification

Code for our paper Multi Scale Curriculum CNN for Context-Aware Breast MRI Malignancy Classification accepted for oral presentation at MICCAI 2019.

TL;DR

We propose a 3D CNN and a multi scale curriculum learning strategy to classify malignancy globally based on an MRI of the whole breast. Our proposed approach does not rely on lesion segmentations, which renders the annotation of training data much more effective than in current object detection approaches. Achieving an AUROC of 0.89, we compare the performance of our approach to Mask R-CNN and Retina U-Net as well as a radiologist. Performance is on par with approaches that, in contrast to our method, rely on pixelwise segmentations of lesions.

License

Copyright (C) 2019 by RWTH Aachen University
http://www.rwth-aachen.de

License: This software is dual-licensed under:

• Commercial license (please contact: [e-mail])
• AGPL (GNU Affero General Public License) open source license

Installation

This repository has been tested with Ubuntu 18.04 and macOS 10.14 Mojave, we highly recommend using a dedicated GPU though.

Editable Installation

git clone https://github.com/haarburger/multi-scale-curriculum.git .
cd multi-scale-curriculum
pip install -e .

Installation

pip install git+https://github.com/haarburger/multi-scale-curriculum.git

Getting Started

A toy example is provided in experiments/example. This folder contains a script and config file for stage 1 and stage 2. You may want to change the folder to which the generated data is saved in experiments/exmple/generate_data.py and the location for the files generated during training in experiments/example/patch.yml and experiments/example/wic.yml. Note that you need to train stage 1 and load the corresponding model during stage 2.

The specific steps are as follows:

# generate dummy data
python experiments/example/generate_data.py

# train stage 1 (3D "patches")
python experiments/example/exec_patch.py -cp experiments/example/patch.yml

# update pretrained model path from the previous step in experiments/example/wic.yml

# train stage 2 (3D volumes)
python experiments/example/exec_wic.py -cp experiments/example/wic.yml

Citation

If you use our work, please cite our paper as follows

@inproceedings{haarburger_2019,
    Author = {Christoph Haarburger and Michael Baumgartner and Daniel Truhn and Mirjam Broeckmann and Hannah Schneider and Simone Schrading and Christiane Kuhl and Dorit Merhof},
    Booktitle = {Medical Image Computing and Computer Assisted Intervention (MICCAI)},
    Doi = {10.1007/978-3-030-32251-9_54},
    Title = {Multi Scale Curriculum CNN for Context-Aware Breast MRI Malignancy Classification},
    Year = {2019}}