This repository contains the code for the term project of UIUC CS 598 DLH, Spring 2023, Team 53. The project is based on the paper Context-aware Health Event Prediction via Transition Functions on Dynamic Disease Graphs (Chang et. al., 2022). The goal of this project is to reproduce the original experiments that predict the health events of patients based on their medical history and context information. We also selected the CGL model (Chang et. al., 2021) to reproduce as the baseline comparison.
- Sherry Li ([email protected])
- Wei-Lun Tsai ([email protected])
- Chang Lu ([email protected])
- Tian Han ([email protected])
- Yue Ning ([email protected])
- Chang Lu
- Chandan K. Reddy
- Prithwish Chakraborty
- Samantha Kleinberg
- Yue Ning
![image](https://private-user-images.githubusercontent.com/28876888/237783805-dbaed888-a7de-4419-943e-bc95c489287f.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.XpeKkb0Cq3JNkNe8b44H5-9TuVAe0v1oDeqM71eSAuw)
git clone https://github.com/willtsai/dlh-sp23-team53.git
cd dlh-sp23-team53
The data can be downloaded from https://mimic.physionet.org/. You will need to obtain permissions for downloading the MIMIC-III (Johnson et. al., 2016) and MIMIC-IV (Johnson et. al., 2023) datasets. Once you have acquired the necessary credentials, you can download the data into the required directories by running the following commands and script:
export PHYSIONET_USERNAME=yourusername
export PHYSIONET_PASSWORD=yourpassword
bash initialize.sh
Dependencies for both our main repro model (Chet) and the baseline model (CGL) are captured in requirements.txt
. To install the dependencies, run the following command:
pip install -r requirements.txt
Run each cell in order from the Jupyter notebooks chet.ipynb
and cgl.ipynb
to preprocess and load the data, train the models, and evaluate accuracy for diagnosis and heart failure prediction tasks.
Alistair Johnson, Tom Pollard, and Roger Mark. 2016. Mimic-iii clinical database.
Alistair Johnson, Lucas Bulgarelli, Tom Pollard, Steven Horng, Leo Anthony Celi, and Roger Mark. 2023. Mimic-iv.
Chang Lu, Tian Han, and Yue Ning. 2022. Context- aware health event prediction via transition functions on dynamic disease graphs. Proceedings of the AAAI Conference on Artificial Intelligence, 36(4):4567โ4574.
Chang Lu, Chandan K Reddy, Prithwish Chakraborty, Samantha Kleinberg, and Yue Ning. 2021. Collabo- rative graph learning with auxiliary text for temporal event prediction in healthcare. In Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21, pages 3529โ3535. International Joint Conferences on Artificial Intelligence Organization. Main Track.