pipeline_tag: image-classification
CheXGenBench: Patient Re-Identification Network
This repository contains the Patient Re-Identification Network (RPN) based on a ResNet-50 architecture, as presented in the paper CheXGenBench: A Unified Benchmark For Fidelity, Privacy and Utility of Synthetic Chest Radiographs.
The model is trained to identify if two chest radiographs (X-rays) belong to the same patient. Within the CheXGenBench framework, it serves as a key component for evaluating the privacy risks and clinical utility of synthetic chest radiograph generation by assessing whether generative models are memorizing or reproducing training data identities.
Links
- Paper: CheXGenBench: A Unified Benchmark For Fidelity, Privacy and Utility of Synthetic Chest Radiographs
- Project Page: https://raman1121.github.io/CheXGenBench/
- GitHub Repository: https://github.com/Raman1121/CheXGenBench
Usage
This model is intended to be used as part of the CheXGenBench evaluation suite to calculate privacy and patient re-identification metrics. For detailed instructions on environment setup and running evaluation scripts (such as privacy_metrics.sh), please refer to the GitHub repository.
Citation
@article{dutt2025chexgenbench,
title={CheXGenBench: A Unified Benchmark For Fidelity, Privacy and Utility of Synthetic Chest Radiographs},
author={Dutt, Raman and Sanchez, Pedro and Yao, Yongchen and McDonagh, Steven and Tsaftaris, Sotirios A and Hospedales, Timothy},
journal={arXiv preprint arXiv:2505.10496},
year={2025}
}