Related Experiment Video
Updated: Oct 2, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Attention-based 3D CNN with residual connections for efficient ECG-based COVID-19 detection
Nebras Sobahi1, Abdulkadir Sengur2, Ru-San Tan3
1King Abdulaziz University, Department of Electrical and Computer Engineering, Jeddah, Saudi Arabia.
Computers in Biology and Medicine
|February 26, 2022
Summary
A new deep learning model using electrocardiography (ECG) can detect COVID-19 with high accuracy. This artificial intelligence approach offers a rapid and accessible method for identifying the viral infection.
Area of Science:
- Artificial Intelligence
- Cardiology
- Infectious Diseases
Background:
- The COVID-19 pandemic has caused millions of deaths globally.
- Pneumonia, a common COVID-19 complication, is diagnosed via imaging.
- Electrocardiography (ECG) can detect cardiovascular injury associated with COVID-19.
Purpose of the Study:
- To develop a deep learning model for COVID-19 detection using ECG data.
- To assess the model's accuracy in binary and multiclass classifications.
Main Methods:
- A novel 19-layer attention-based 3D convolutional neural network (CNN) with residual connections was designed.
- The model processed 12-lead ECG data from normal subjects, COVID-19 patients, and those with abnormal heartbeats.
- ECG data was preprocessed into 12-dimensional volumes for network input.
Main Results:
- The model achieved 99.0% accuracy for binary classification (COVID-19 vs. normal).
- Multiclass classification accuracy reached 92.0% across all groups.
- Ten-fold cross-validation was employed to evaluate performance.
Conclusions:
- The developed deep learning model demonstrates high efficacy for COVID-19 detection using ECG.
- This approach offers a promising, accessible tool for rapid COVID-19 screening.
- The model is prepared for validation on larger ECG databases.
