ventmode

Reproducible, Accurate, Ventilator Mode Classifications using PB-840 Ventilator Data

Citing

If you used my dataset or the work herein please cite us

@article{rehm2019improving,
  title={Improving Mechanical Ventilator Clinical Decision Support Systems with a Machine Learning Classifier for Determining Ventilator Mode.},
  author={Rehm, Gregory B and Kuhn, Brooks T and Nguyen, Jimmy and Anderson, Nicholas R and Chuah, Chen Nee and Adams, Jason Yeates},
  journal={Studies in health technology and informatics},
  volume={264},
  pages={318--322},
  year={2019}
}

Install

Clone the repo

git clone https://github.com/hahnicity/ventmode

You should use anaconda for installing dependencies. Install anaconda then create a new environment.

cd ventmode
conda env create -f environment.yml
pip install -e .

Reproducing Work

You can reproduce the work we've done by running a bash script.

git checkout reproduction_tag
cd ventmode
./experiments.sh

Using This In Your Own Projects

We've designed this to be as easy as possible to use in your own projects. There is an API that is capable of being used and there is also a CLI script. Let's start with the CLI. First you will need to save an instance of the trained Random Forest model, and the data scaler. We recommend that you train your model on a 20% split of the data because we have found this leads to best subsequent model performance.

cd ventmode
python main.py --save-classifier-to saved-ventmode-model.pkl -sr .2 --split-type simple

Then you can use the CLI by

python scan_file.py <your filename> saved-ventmode-model.pkl saved-ventmode-model.pkl.scaler

For API usage:

    from ventmode import datasets
    from ventmode.main import run_dataset_with_classifier

    fileset = {'x': [('patient1', <patient 1's file>), ('patient 2', <patient 2's file), ...]}
    vfinal = datasets.VFinalFeatureSet(fileset, 10, 100)
    df = vfinal.create_prediction_df()
    cls = pickle.load(open(<path to saved classifier>))
    scaler = pickle.load(open(<path to saved scaler>))
    model_results = run_dataset_with_classifier(cls, scaler, df, "vfinal")

ShixuanGu/ventmode

ventmode

Citing

Install

Reproducing Work

Using This In Your Own Projects