willxxy / Text-EGM

[CHIL 2024] Interpretation of Intracardiac Electrograms Through Textual Representations

Interpretation of Intracardiac Electrograms Through Textual Representations

William Jongwon Han, Diana Gomez, Avi Alok, Chaojing Duan, Michael A. Rosenberg, Douglas Weber, Emerson Liu, Ding Zhao.

Official code for "Interpretation of Intracardiac Electrograms Through Textual Representations" accepted by 2024 Conference on Health, Inference, and Learning (CHIL).

If you experience any bugs or have any questions, please submit an issue or contact at wjhan{at}andrew{dot}cmu{dot}edu.

We thank the Mario Lemieux Center for Heart Rhythm Care at Allegheny General Hospital for supporting this work.

Set Up Environment

Note: We have only tested on Ubuntu 20.04.5 LTS.

1. conda create -n envname python=3.8

2. conda activate envname

3. git clone https://github.com/willxxy/ekg-af.git

4. cd ekg-af

5. pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu113

Note: Please ensure that the pip you are using is from the conda environment

6. Test if pytorch version is compatible with current, available gpus by executing python gpu.py. Currently, we have only tested on A5000 (24 GB) and A6000 (48 GB) NVIDIA GPUs.

7. pip install -r requirements.txt

Set Up Data

1. Although the data we curated is not publicly available, we do have experimental results on an external dataset (main results are in Table 2 in the paper), namely the "Intracardiac Atrial Fibrillation Database" available on PhysioNet.

2. To set up this data, cd into the preprocess folder.

3. Please execute the following command to download the data.

wget https://physionet.org/static/published-projects/iafdb/intracardiac-atrial-fibrillation-database-1.0.0.zip

4. Unzip the file by executing

unzip intracardiac-atrial-fibrillation-database-1.0.0

5. Now execute the folllowing command to preprocess the data.

sh preprocess.sh

6. This should create a data folder with several .npy for training, validation, and test.

Start Training

1. From the preprocess folder cd ../ back to the main directory.

2. You can now directly use train.sh files to start training.

Inference

1. Please execute sh inference.sh after training. Make sure to specify the checkpoint path.

Visualizations

All visualizations will be saved under their respective checkpoint folder. Please cd visualize before visualizing. Under the visualize folder, please view the following scripts:

1. stitch.sh - Visualizes the reconstructed and forecasted signals.

2. viz_tokens.sh - Visualizes the tokenized representation of the signal.

3. viz_attentions.sh - Visualizes the attention map of the model.

4. viz_int_grad.sh - Visualizes the attribution scores of the model.

Citation

If you found this repository or work helpful to your own, please cite the following bibtex.

@misc{han2024interpretation,
      title={Interpretation of Intracardiac Electrograms Through Textual Representations}, 
      author={William Jongwon Han and Diana Gomez and Avi Alok and Chaojing Duan and Michael A. Rosenberg and Douglas Weber and Emerson Liu and Ding Zhao},
      year={2024},
      eprint={2402.01115},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}