challenge2012

Python code parsing data from PhysioNet Challenge 2012.

The PhysioNet 2012 challenge, described here, focused on encouraging development of mortality prediction models for intensive care unit patients. Time-stamped observations for a number of measurements were made available, with the goal to predict in-hospital mortality.

A big part of any machine learning project is data preprocessing. The notebook prep-data.ipynb transforms the original PhysioNet 2012 data to have one row per patient. This is done by applying a number of simple feature extraction steps to each measurement time-series independently, e.g. extracting the first heart rate, first blood pressure, and so on. The code also pre-processes the data to remove outliers based on physiologic limits (you cannot have a negative blood pressure, you cannot have an oxygen saturation above 100%, and so on).

The logic used here was applied in our conference paper: Patient specific predictions in the intensive care unit using a Bayesian ensemble. Our entry had the highest score 1 in the competition, where score 1 was defined as the minimum of the sensitivity/positive predictive value.

The code was originally written in MATLAB but with any luck this Python code is a faithful reproduction.

Downloading the data

For convenience, the already processed datasets are made available in the data subfolder.

If you use this data in your research or otherwise, please do acknowledge the organizers of the PhysioNet 2012 Challenge:

@article{silva2012predicting,
  title={Predicting in-hospital mortality of icu patients: The physionet/computing in cardiology challenge 2012},
  author={Silva, Ikaro and Moody, George and Scott, Daniel J and Celi, Leo A and Mark, Roger G},
  journal={Computing in cardiology},
  volume={39},
  pages={245},
  year={2012},
  publisher={NIH Public Access}
}

We would also appreciate acknowledgement:

@inproceedings{johnson2012patient,
  title={Patient specific predictions in the intensive care unit using a Bayesian ensemble},
  author={Johnson, Alistair EW and Dunkley, Nic and Mayaud, Louis and Tsanas, Athanasios and Kramer, Andrew A and Clifford, Gari D},
  booktitle={Computing in Cardiology (CinC), 2012},
  pages={249--252},
  year={2012},
  organization={IEEE}
}