Mechanical Ventilation Clustering
The advancement of Electronic Health Records (EHRs) and machine learning have enabled a data-driven and personalised approach to healthcare. One step in this direction is to uncover patient sub-types with similar disease trajectories in a heterogeneous population. This is especially important in the context of mechanical ventilation in intensive care, where mortality is high and there is no consensus on treatment. In this work, we present an approach to clustering mechanical ventilation episodes, using a multi-task combination of supervised, self-supervised and unsupervised learning techniques. Our dynamic clustering assignment is guided to reflect the phenotype, trajectory and outcomes of the patient. Experimentation on a real-world dataset is encouraging, and we hope that this could translate into actionable insights in guiding future clinical research.
To run the sql files you must have the AmsterdamUMCdb database set up: https://amsterdammedicaldatascience.nl/amsterdamumcdb.
Follow the instructions: https://github.com/AmsterdamUMC/AmsterdamUMCdb to ensure the correct connection configuration.
Replace the amsterdam_path in
paths.jsonto a convenient location in your computer, and do the same for
amsterdamUMCdb_preprocessing/create_all_tables.sqlusing find and replace for
'/Users/emmarocheteau/PycharmProjects/Mechanical-Ventilation-Clustering/amsterdam_data/'. Leave the extra ‘/’ at the end.
In your terminal, navigate to the project directory, then type the following commands:
psql 'dbname=amsterdam user=amsterdam options=--search_path=amsterdam'
Inside the psql console:
This step might take a couple of hours.
To quit the psql console:
Then run the pre-processing scripts in your terminal. This will need to run overnight:
python3 -m amsterdamUMCdb_preprocessing.run_all_preprocessing
It will create the following directory structure:
MIMIC_data ├── test │ ├── data.npz │ └── vents.txt ├── train │ ├── data.npz │ └── vents.txt ├── val │ ├── data.npz │ └── vents.txt ├── flat_features.csv ├── labels.csv ├── standardisation_limits.csv ├── timeseries.csv ├── ventilator_settings.csv └── vents.txt
Running the models
Once you have run the pre-processing steps you can run all the models in your terminal. Set the working directory to Mechanical-Ventilation-Clustering, and run the following:
python3 -m train_prediction_only
python3 -m train_prediction_reconstruction
Note that your experiment can be customised by using command line arguments (these can be explored in args.py).
Each experiment you run will create a directory within logs. The naming of the directory is based on the date and time that you ran the experiment (to ensure that there are no name clashes).