Machine Learning for Cardiac Electrical Imaging

Welcome to the GitHub repository for machine learning in cardiac electrical imaging! This repository contains code specifically tailored for the research paper and dataset described below:

• (Paper) M. Landajuela, R. Anirudh, J. Loscazo, and R. Blake, "Intracardiac Electrical Imaging Using the 12-Lead ECG: A Machine Learning Approach Using Synthetic Data," presented at the 2022 Computing in Cardiology (CinC) conference in Tampere, Finland. The paper can be accessed here. It is published in the proceedings of CinC 2022, pages 1-4. The DOI for this paper is 10.22489/CinC.2022.026.

• (Dataset) M. Landajuela, R. Anirudh, and R. Blake have curated a comprehensive dataset of simulated intracardiac transmembrane voltage recordings and ECG signals. This dataset is part of the Lawrence Livermore National Laboratory (LLNL) Open Data Initiative and can be accessed here. The dataset was published in the UC San Diego Library Digital Collections in 2022.

• (Blog Post) M. Landajuela, Machine Learning for Cardiac Electrocardiography here.

The code in this repository supports two primary tasks:

1. Activation Map Reconstruction from ECG
2. Transmembrane Potential Reconstruction from ECG

Feel free to explore the code and utilize it for your own research or projects related to cardiac electrical imaging.

Overview

Installation

To install the code, you can use the following commands:

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Code Structure

The main code is in the following files:

File	Description
trainable_model.py	Base class for all the models
io_util.py	Utility functions for input/output
ml_util.py	Utility functions for machine learning
data_interface.py	Interface to the data
learn_ecg2time.py	Main code for task 1
learn_ecg2vm.py	Main code for task 2

Configuration files

Configuration files can be found in config/ecg2*.config. For example, the configuration file for ecg2time is ecg2time.config. The configuration files contain the following sections:

Section	Description
[DATA]	Data related parameters
[MODEL]	Model related parameters
[TRAIN]	Training related parameters
[SAVE]	Logging related parameters

Run

To run the code, you can use the following commands:

python3 learn_ecg2time.py config/ecg2time.config
python3 learn_ecg2vm.py config/ecg2vm.config

Note: You can create an input deck directory for the training ecg2*_input_deck and run the code from the input deck directory.

python3 ../learn_ecg2time.py ../config/ecg2time.config
python3 ../learn_ecg2vm.py ../config/ecg2vm.config

Post-process

To post-process the results, you can use the following commands:

python3 ./tools/read_training_stats.py trainingStats_errors.h5 -plot True

Visualization

We provide a typical heart mesh and the corresponding recording locations in the geometry folder. The mesh is in the .vtk format and the recording locations are in the .csv format. You can visualize the mesh and the recording locations using ParaView. Note that the recording locations are in the same coordinate system as the mesh. Also, the indices of the recorded transmembrane potentials correspond to the indices of the recording locations. One way to interpolate the predicted values at the recording locations to the whole cardiac mesh is to use the ShepardKernel filter in ParaView.

Data

The data associated with this code is provided in the following repository: Dataset of Simulated Intracardiac Transmembrane Voltage Recordings and ECG Signals.

Steps to download the dataset (Click to expand)

To download the data, you can use the following command:

source download_intracardiac_dataset.sh

Once, it is downloaded, you can point to the data using the datapaths_train and datapaths_val in the configuration file. For example, the configuration file for ecg2time looks like:

[DATA]
datapaths_train = [full path of intracardiac_dataset]/data_hearts_dd_0p2
datapaths_val = [full path of intracardiac_dataset]/data_hearts_dd_0p2

Note: You might want to change the train and validation data path to point to your split of the data.

License

cardiac_ml is distributed under the terms of the MIT license.

All new contributions must be made under the MIT license.

See LICENSE, and NOTICE for details.

LLNL-CODE-850741