search

seorim0 / ResUNet-LC

2D residual U-Net (ResUNet) and a lead combiner (LC) for 12-lead ECG Abnormality Classification
MIT License
10 stars 0 forks source link
abnormal-detection abnormality-detection classification deep-learning deep-neural-networks dnn ecg ecg-classification electrocardiogram icassp icassp2024 multi-label-classification pytorch resnet

readme

MULTI-LABEL ABNORMALITY CLASSIFICATION FROM 12-LEAD ECG USING A 2D RESIDUAL U-NET

This is an official repo of the paper "MULTI-LABEL ABNORMALITY CLASSIFICATION FROM 12-LEAD ECG USING A 2D RESIDUAL U-NET," which is accepted to ICASSP 2024.

Abstract：This paper proposes a two-dimensional (2D) deep neural network (DNN) model for the electrocardiogram (ECG) abnormality classification, which effectively utilizes the inter and intra-lead information comprised in the 12-lead ECG. The proposed model is designed using a stack of residual U-shaped (ResU) blocks so that it can effectively capture ECG features in a multiscale. The 2D features extracted by the ResU block are down-mixed to 1D features using a lead combiner block designed to merge features of the lead domain into both the time and channel domain. Through experiments, we confirm that our model outperforms other state-of-the-art models in various metrics.

Update:

Requirements

This repo is tested with Ubuntu 22.04, PyTorch 2.0.1, Python3.10, and CUDA11.7. For package dependencies, you can install them by:

pip install -r requirements.txt

Getting started

  1. Install the necessary libraries.

  2. Download the PhysioNet Challenge 2021 database and place it in '../Dataset/' folder. 

    ├── 📦 ResUNet_LC   
│   └── 📂 dataset   
│       └── 📜 train_dataset.csv   
│       └── 📜 test_dataset.csv   
│   └── ...   
└── 📦 Dataset   
└── 📂 physionet_challenge_dataset
    └── 📂 physionet.org 
        └── ...

    If you want to get csv file, please contact us.

  3. Run train_interface.py

    • You can simply change any parameter settings if you need to adjust them. (options.py)

Results

001

002

f1_scores_graph_distribution

auprc_graph_distribution

Reference

Will Two Do? Varying Dimensions in Electrocardiography: The PhysioNet/Computing in Cardiology Challenge 2021
Matthew Reyna, Nadi Sadr, Annie Gu, Erick Andres Perez Alday, Chengyu Liu, Salman Seyedi, Amit Shah, and Gari Clifford
[paper]
Automatic diagnosis of the 12-lead ECG usinga deep neural network
Antônio H. Ribeiro, et al.
[paper] [code]
A multi-view multi-scale neural network for multi-label ECG classification
Shunxiang Yang, Cheng Lian, Zhigang Zeng, Bingrong Xu, Junbin Zang, and Zhidong Zhang
[paper] [code]
Classification of ECG using ensemble of residual CNNs with attention mechanism
Petr Nejedly, Adam Ivora, Radovan Smisek, Ivo Viscor, Zuzana Koscova, Pavel Jurak, and Filip Plesinger
[paper] [code]

Contact

Please get in touch with us if you have any questions or suggestions.
E-mail: allmindfine@yonsei.ac.kr (Seorim Hwang) / jbcha7@yonsei.ac.kr (Jaebine Cha)