d909b / heart_rhythm_attentive_rnn

❤️📱 Heart rhythm classification from mobile event recorder data using attentive neural networks.

Beat by Beat: Classifying Cardiac Arrhythmias with Recurrent Neural Networks

Contact: Patrick Schwab, ETH Zurich [email protected]

Authors: See AUTHORS.txt

License: GPLv3; See LICENSE.txt

Description: Predicts the rhythm of given ECG signals using ensembles of recurrent neural networks. We delineated our approach in this manuscript. This solution is an entry to the PhysioNet / CinC challenge 2017.

Citation

If you reference our methodology, code or results in your work, please consider citing:

@inproceedings{schwab2017beat,
  title={Beat by Beat: Classifying Cardiac Arrhythmias with Recurrent Neural Networks},
  author={Schwab, Patrick and Scebba, Gaetano and Zhang, Jia and Delai, Marco and Karlen, Walter},
  booktitle={Computing in Cardiology},
  year={2017}
}

Installation

Requires:

• pip
• Keras >= 1.2.2
• Theano >= 0.8.2
• matplotlib >= 1.3.1
• pandas >= 0.18.0
• h5py >= 2.6.0
• scikit-learn == 0.17.1
• pywavelets == 0.2.2
• imbalanced_learn == 0.2.1
• pyhsmm == 0.1.7

To train models you need to download the PhysioNet 2017 challenge data.

ATTENTION - PATCHES REQUIRED:

To save bidirectional RNNs you need to additionally patch the version of Keras installed by pip using this patch.

To train HSMMs you need to patch the PyHSMM library in pyhsmm_states.py line 1071 to: obs, offset = obs[:,state], offset