ECG Real-time Arrhythmia Monitoring

Code files to accompany the paper "ECG-based Real-time Arrhythmia Monitoring Using Quantized Deep Neural Networks: A Feasibility Study". For more details about this study, please visit: https://intsav.github.io/realtime_ecg.html

Step 1 - Requirements

Clone this repository

  git clone [email protected]:intsav/RealtimeArrhythmiaMonitoring

Install virtualenv

  pip3 install virtualenv

Create and activate Python 3.7 environment

  virtualenv -p python3.7

Install requirements

  ./setup.sh

Step 2 - Data

In the root directory create a new directories called model and data

  mkdir {models,data}

Fetch and save dataset

wget --load-cookies /tmp/cookies.txt "https://docs.google.com/uc?export=download&confirm=$(wget --quiet --save-cookies /tmp/cookies.txt --keep-session-cookies --no-check-certificate 'https://docs.google.com/uc?export=download&id=1KIBxRB12tbEop02Dj_sLBuZvPgu3ua6e' -O- | sed -rn 's/.*confirm=([0-9A-Za-z_]+).*/\1\n/p')&id=1KIBxRB12tbEop02Dj_sLBuZvPgu3ua6e" -O data/mitdb_360_train.csv  && rm -rf /tmp/cookies.txt
  wget --no-check-certificate 'https://docs.google.com/uc?export=download&id=1epF6BHCrTUOrpILBUp4xg160guVy_Jsr' -O data/mitdb_360_test.csv

Step 3 - Training

Run the following command from the root directory

python3 code/train.py

At each epoch the model is saved in models directory.

Step 4 - Test baseline model

Run the following command from the root directory by selecting your best model

python3 code/test.py --model models/**FILENAME**

Step 5 - Test quantized model

Download quantized weights

  wget --no-check-certificate 'https://docs.google.com/uc?export=download&id=1pY--6B4xNpcEMixEEwVgoD1h5AVvk2pW' -O models/ecg_quant.tflite

Test quantized model

python3 code/test_quant.py

License

This project is licensed under the GNU General Public License v3.0 - see the LICENSE file for details.