gaiamrc / ecgtransformer

Source code of BIBM 2019 Paper "Fusing Transformer Model with Temporal Features for ECG Heartbeat Classification"

Acknowledgement

Some of the python code (comment lines or functions) in this project is not used which are some attempts or hints I have done. I must admit that the code refers to the following two items in many places:

1. https://github.com/jadore801120/attention-is-all-you-need-pytorch
2. https://github.com/eladhoffer/seq2seq.pytorch

MATLAB code

version is R2018.a

1. Need WFDB Toolbox for MATLAB from https://physionet.org/physiotools/matlab/wfdb-app-matlab/
2. run get_anno to get annotation
3. run denoising to get denoised signal
4. run segmentation to segment the denoised signal
5. run features to get the final results, format: label, preRR interval, postRR interval, [heartbeat signal]
6. Here I provide a version of the results, you can also try to use different noise reduction methods or segmentation methods.

Python code

1. All parameters are defined in config.py
2. Some package versions：
   • python: 3.6.8
   • numpy: 1.16.0
   • Pytorch: '1.0.1.post2'
   • cuda: '10.0.130'
   • tqdm: 4.31.1
3. You may need to change the code in main.py to set train_file and valid_file path according to your preference.
4. Since I have modified my code, I'm not sure the code will work correctly. So if there is any bug, please let me know.

License

For academtic and non-commercial usage