UAE

Implementation of "Abnormality Detection in Chest X-ray Images Using Uncertainty Prediction Autoencoders" in MICCAI 2020. [paper]

Datasets

RSNA Pneumonia Detection Challenge Dataset: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/data

Pediatric Dataset: https://data.mendeley.com/datasets/rscbjbr9sj/3

Requirements

• python, pytorch, numpy, scikit-learn, tensorboard-pytorch

Basic usage

Before running the scripts, you should edit DATA_PATH in xray_data.py in order to read the data correctly:

DATA_PATH = '$WHERE_YOU_STORE_DATASETS'

To train a traditional autoencoder:

python uae_main.py

To train an autoencoder with uncertainty:

python uae_main.py --u

To evaluate a model:

python uae_main.py --eval --u(optional)

Cite

@inproceedings{mao2020abnormality,
  title={Abnormality Detection in Chest X-Ray Images Using Uncertainty Prediction Autoencoders},
  author={Mao, Yifan and Xue, Fei-Fei and Wang, Ruixuan and Zhang, Jianguo and Zheng, Wei-Shi and Liu, Hongmei},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={529--538},
  year={2020},
  organization={Springer}
}