AoOR: Leveraging Interpretable Machine Learning in Intensive Care

Overview

This repository supports the research paper "Leveraging Interpretable Machine Learning in Intensive Care," which explores the use of interpretable machine learning models in ICU settings to aid medical professionals in decision-making.

Key Findings

Our research challenges the prevailing belief that only black-box models can provide high predictive performance in healthcare. We demonstrate that:

Interpretable models can achieve competitive predictive performance:
- Only a minor decrease of 0.2-0.9 percentage points in AUROC compared to black-box models
- Full interpretability maintained
Parsimonious models are highly effective:
- Utilize only 2.2% of available patient features
- Negligible performance drop relative to black-box models:
  - Range: 0.1 to 1.0 percentage points
  - Average: 0.5 percentage points

These findings aim to inspire further research and development of interpretable ML models in healthcare applications, showcasing that accuracy need not be sacrificed for interpretability.

Citation

If you use this work in your research, please cite:

@article{
  title={Leveraging Interpretable Machine Learning in Intensive Care},
  author={Bohlen, Lasse and Rosenberger, Julian and Zschech, Patrick and Kraus, Mathias},
  journal={[Journal Name]},
  year={[Year]},
  volume={[Volume]},
  pages={[Pages]},
  doi={[DOI]}
}

Prerequisites

Access to the MIMIC-III database via PhysioNet
Conda for managing the environment

Installation

Step 1: Export MIMIC-III Data

Obtain access to the MIMIC-III database via PhysioNet.
Follow the instructions provided in the edited version of the MIMIC-III benchmark suite to export the data in CSV format.
Save the exported data as mimic_complete.csv in the data/raw/MIMIC/ directory.

Step 2: Set Up the Repository

Clone this repository into your workspace:

git clone https://github.com/username/XXX.git
cd Interpretable_ICU_predictions

Create and activate the conda environment:

conda env create -f env.yml
conda activate envMIMIC

Usage

Run the Python scripts found in the scripts/ directory to execute the experiments. For example:

python scripts/experiment_X.py

Experiments

For example one experiment could be:

experiment_X.py: Mortality predicition
...

Structure

Interpretable_ICU_predictions/
├── data/
│   └── raw/
│       └── MIMIC/
│           └── mimic_complete.csv
├── hpo_configs/
├── logging/
├── output/
├── results/
├── scripts/
│   ├── classes/
│   ├── utils/
│   ├── experiment_X.py
│   ├── experiment_Y.py
│   └── ...
├── env.yml
├── README.md
└── ...

Contributing

We welcome contributions to this project. Please follow these steps:

Fork the repository
Create a new branch(git checkout -b feature-branch)
Make your changes and commit (git commit -am 'Add some feature')
Push to the branch (git push origin feature-branch)
Create a new Pull Request

License

This project is licensed under the MIT license.

Contact

For questions or feedback, please contact Lasse Bohlen at [lasse.bohlen@fau.de].

Acknowledgments

MIMIC-III Database

@article{johnson2016mimic,
  title={MIMIC-III, a freely accessible critical care database},
  author={Johnson, Alistair EW and Pollard, Tom J and Shen, Lu and Lehman, Li-wei H and Feng, Mengling and Ghassemi, Mohammad and Moody, Benjamin and Szolovits, Peter and Anthony Celi, Leo and Mark, Roger G},
  journal={Scientific data},
  volume={3},
  number={1},
  pages={1--9},
  year={2016},
  publisher={Nature Publishing Group}
}

MIMIC-III Benchmarks

@article{Harutyunyan2019,
  author={Harutyunyan, Hrayr and Khachatrian, Hrant and Kale, David C. and Ver Steeg, Greg and Galstyan, Aram},
  title={Multitask learning and benchmarking with clinical time series data},
  journal={Scientific Data},
  year={2019},
  volume={6},
  number={1},
  pages={96},
  issn={2052-4463},
  doi={10.1038/s41597-019-0103-9},
  url={https://doi.org/10.1038/s41597-019-0103-9}
}

HB-Dynamite/Interpretable_ICU_predictions