ECG classifier code for Python for the PhysioNet/CinC Challenge 2020

This is a PyTorch implementation of "Multi-label Classification of Electrocardiogram with Modified Residual Networks" paper.

Contents

This code uses two main scripts to train the model and classify the data:

• train_model.py Train your model. Add your model code to the train_12ECG_model function. It also performs all file input and output. Do not edit this script or we will be unable to evaluate your submission.
• driver.py is the classifier which calls the output from your train_model script. It also performs all file input and output. Do not edit this script or we will be unable to evaluate your submission.

Check the code in these files for the input and output formats for the train_model and driver scripts.

To create and save your model, you should edit train_12ECG_classifier.py script. Note that you should not change the input arguments of the train_12ECG_classifier function or add output arguments. The needed models and parameters should be saved in a separated file. In the sample code, an additional script, get_12ECG_features.py, is used to extract hand-crafted features.

To run your classifier, you should edit the run_12ECG_classifier.py script, which takes a single recording as input and outputs the predicted classes and probabilities. Please, keep the formats of both outputs as they are shown in the example. You should not change the inputs and outputs of the run_12ECG_classifier function.

Use

You can run this classifier code by installing the requirements and running

python train_model.py training_data model   
python driver.py model test_data test_outputs

where training_data is a directory of training data files, model is a directory of files for the model, test_data is the directory of test data files, and test_outputs is a directory of classifier outputs. The PhysioNet/CinC 2020 webpage provides a training database with data files and a description of the contents and structure of these files.

Submission

The driver.py, get_12ECG_score.py, and get_12ECG_features.py scripts must be in the root path of your repository. If they are inside a folder, then the submission will be unsuccessful.

Details

See the PhysioNet/CinC 2020 webpage for more details, including dataset and instructions for the other files in this repository.

Transfer learning using pre-trained

The dataset for pre-trained included 40,000 medical ecg samples provided by the Engineering Research Center of the Ministry of Education for mobile Health Management System of Hangzhou Normal University, China. Each sample has 8 leads, namely I, II, V1, V2, V3, V4, V5, and V6. You can also calculate the data of the remaining 4 leads by using the following formula:

III=II-I aVR=-(I+II)/2 aVL=I-II/2 aVF=II-I/2

Each sample was sampled at a frequency of 500 HZ, a length of 10 seconds, and a unit voltage of 4.88 microvolts.

code for pre-traind is comming soon.

Docker command

1、docker build -t image:v1.0 .

2、nvidia-docker run -it -v /home/yangshan/cinc2020_pytorch/data:/physionet/data -v /home/yangshan/cinc2020_pytorch/output:/physionet/output -v /home/yangshan/cinc2020_pytorch/test_data:/physionet/test_data -v /home/yangshan/cinc2020_pytorch/model:/physionet/model image:v1.0 bash

3、python train_model.py data/ model/

4、python driver.py model/ data/ output/

5、python evaluate_12ECG_score.py data/ output/