/rrs_public

Rapid Response System Public

Primary LanguageJupyter Notebook

rrs_public

Rapid Response System Public code

Predict patient's in-hospital abnormal status for rapid response system

Overall system

  • Input: Patient's clinical variables (includes vital signs and laboratory tests) in an observation peroid
  • Output: Patient's abnormal status/ abnormal probability

overall_dews

Window interval processing

  • Window time D contains the measurement features of patient in n timepoints (in the figure, n = 8)
  • The window time D will slides for every timepoint for each patient
  • With n = 8, for each patient, we should wait 8 hours for the first output
  • After that, the system will predict once every hour

Overall

  • Proposed model: Temporal Variational Autoencoder (TVAE)

tvae

Running:

  • Install enviroment dependencies via 'requirement.txt'
  pip install -r requirement.txt
  • Install torch:
conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
  • Configure the file path on the functions
  • Run 'CNUH_prediction.py' for predict the outcome