KANNEDS - Using Kinematics Analysis and Neural Networks for Early Detection of Sepsis

These are the LSTM Neural Network source code and a sample dataset of the KANNEDS project.

The code uses the files below and shows a 5-fold cross validation of the LSTM neural network with the Kinematics Features (KF) as input and with Vital Signs (VS) as input:

• KANNEDS_NEGATIVES_SAMPLE_DATA_FINAL.csv: Sample dataset containing 1431 patients negative for sepsis
• KANNEDS_POSITIVES_SAMPLE_DATA_FINAL.csv: Sample dataset containing 877 patients positive for sepsis
• KANNEDS_VARIABLES_FINAL.csv: Table containing the clinical variables used.

To run the code, please proceed as follow:

1. Download the files keeping the folder Input;
2. Install a Python v3.6.9 environment with Keras v2.2.4 and Tensorflow v1.12 packages;
3. Run the Python code.