aehrc/cxrmate-tf

CXRMate: Longitudinal Data and a Semantic Similarity Reward for Chest X-Ray Report Generation

This is the longitudinal, multi-image Chest X-Ray (CXR) report generator trained with teacher forcing (no reinforcement learning).

Hugging Face Models From This Paper:

Link	Longitudinal	Multi-image	Reinforcement learning
https://huggingface.co/aehrc/cxrmate	Y	Y	Y
https://huggingface.co/aehrc/cxrmate-tf	Y	Y	N
https://huggingface.co/aehrc/cxrmate-multi-tf	N	Y	N
https://huggingface.co/aehrc/cxrmate-single-tf	N	N	N

Paper:

• https://doi.org/10.1016/j.imu.2024.101585

@article{NICOLSON2024101585,
title = {Longitudinal data and a semantic similarity reward for chest X-ray report generation},
journal = {Informatics in Medicine Unlocked},
volume = {50},
pages = {101585},
year = {2024},
issn = {2352-9148},
doi = {https://doi.org/10.1016/j.imu.2024.101585},
url = {https://www.sciencedirect.com/science/article/pii/S2352914824001424},
author = {Aaron Nicolson and Jason Dowling and Douglas Anderson and Bevan Koopman},
}

ArXiv:

• https://arxiv.org/abs/2307.09758

GitHub Repository:

• https://github.com/aehrc/cxrmate

Notebook Example:

• https://github.com/aehrc/cxrmate/blob/main/examples/cxrmate-tf.ipynb