This repository provides comprehensive pipelines for generating vector embeddings from MIMIC III and MIMIC IV datasets, as well as related healthcare datasets. The primary focus is on creating high-quality vector representations that can be used for downstream machine learning tasks in healthcare AI research.
- Multi-dataset Support: Full compatibility with MIMIC III, MIMIC IV, and related healthcare datasets
- State-of-the-art Models: Integration with advanced embedding models.
-
Register for PhysioNet Access:
- Create an account at PhysioNet
- Complete the required training for human subjects research
- Request access to MIMIC III and/or MIMIC IV datasets
-
Download Datasets:
# Example for MIMIC IV version 3.1 wget -r -N -c -np --user YOUR_USERNAME --ask-password \ https://physionet.org/files/mimiciv/3.1/
- Repository: https://github.com/Stanford-AIMI/CheXagent
- Hugging Face: https://huggingface.co/StanfordAIMI/CheXagent-8b
- Citation: Chen, Z., et al. (2023). "CheXagent: Towards a Foundation Model for Chest X-Ray Interpretation" arXiv preprint arXiv:2401.12208
- Repository: https://github.com/RPIDIAL/CheXFound
- Citation: Yang, Z., Xu, X., Zhang, J., Wang, G., Kalra, M. K., & Yan, P. (2025). Chest X-ray Foundation Model with Global and Local Representations Integration. ArXiv. https://arxiv.org/abs/2502.05142
- Repository: https://github.com/hustvl/EVA-X
- Hugging Face: https://huggingface.co/MapleF/eva_x/tree/main
- Citation: Yao, J., Wang, X., Song, Y., Zhao, H., Ma, J., Chen, Y., Liu, W., & Wang, B. (2024). EVA-X: A foundation model for general chest X-ray analysis with self-supervised learning.
- Documentation: https://developers.google.com/health-ai-developer-foundations/cxr-foundation
- Repository: https://github.com/google-health/cxr-foundation
- Hugging Face: https://huggingface.co/google/cxr-foundation
- Repository: https://github.com/google-health/medsiglip
- Hugging Face: https://huggingface.co/google/medsiglip-448
- Documentation: https://developers.google.com/health-ai-developer-foundations/medsiglip
- Citation: Google Health AI. (2024). MedSigLIP: Medical image understanding with SigLIP.
- Repository: https://github.com/mlmed/torchxrayvision
- Documentation: https://mlmed.org/torchxrayvision/
- Citation: Cohen, J. P., Viviano, J. D., Bertin, P., Morrison, P., Torabian, P., Guarrera, M., Lungren, M. P., Chaudhari, A., Brooks, R., Hashir, M., & Bertrand, H. (2022). TorchXRayVision: A library of chest X-ray datasets and models. Proceedings of Machine Learning for Health, 172, 231-249.
- Repository: https://github.com/echonet/EchoPrime
- Repository: https://github.com/CarDS-Yale/PanEcho
- Repository: https://github.com/Edoar-do/HuBERT-ECG?tab=readme-ov-file
- Hugging Face: https://huggingface.co/Edoardo-BS
- Repository: https://github.com/control-spiderman/ECGFM-KED
- Repository: https://github.com/PKUDigitalHealth/ECGFounder
- Hugging Face: https://huggingface.co/PKUDigitalHealth/ECGFounder/tree/main
- Johnson, A. E. W., et al. "MIMIC-IV, a freely accessible electronic health record dataset." Scientific Data 10.1 (2023): 1.
- Goldberger, A. L., et al. "PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals." Circulation 101.23 (2000): e215-e220.
- Tohyama T, et al. Multi-view echocardiographic embedding for accessible AI development. medRxiv. 2025. doi:10.1101/2025.08.15.25333725.
- Chung DJ, et al. Echocardiogram Vector Embeddings Via R3D Transformer for the Advancement of Automated Echocardiography. JACC Adv 2024;3:101196.
This project is licensed under the MIT License - see the LICENSE file for details.
- Embedding of the MIMIC chest-x-ray, echo, ecg will be made available on physionet.
- Due to process time, this is currently temporarily available at https://drive.google.com/drive/folders/18pOkTlu-LguUZp3TLKqZSjSBky-alhja?usp=sharing
- Issues: GitHub Issues
- Discussions: GitHub Discussions
 |
 |
|---|
- Funding: This research was supported by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2024-00439677)
Disclaimer: This software is provided for research purposes only. It is not intended for clinical use. Always comply with your institution's ethics and data usage policies when working with healthcare data.