Classification of COVID-19 electrocardiograms by using hexaxial feature mapping and deep learning

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Coronavirus disease 2019 (COVID-19) is a global pandemic requiring improved diagnostic methods.
• Current diagnostic techniques for COVID-19 have limitations, necessitating novel approaches.
• Early diagnosis of COVID-19 is critical for patient outcomes and disease management.

Purpose of the Study:

• To propose a novel deep learning framework for automated COVID-19 diagnosis using Electrocardiogram (ECG) data.
• To develop and validate a new hexaxial feature mapping technique for ECG representation.
• To assess the diagnostic performance of the proposed method in classifying COVID-19 cases.

Main Methods:

• A novel hexaxial feature mapping method was developed to convert 12-lead ECG data into 2D colorful images.
• Gray-Level Co-Occurrence Matrix (GLCM) was employed for feature extraction from ECG data.
• A new Convolutional Neural Network (CNN) architecture was designed and utilized for COVID-19 classification based on the generated ECG images.

Main Results:

• The proposed approach achieved 96.20% accuracy and 96.30% F1-Score in classifying COVID-19 versus normal ECG data.
• In a broader classification scenario, the method demonstrated 93.00% accuracy and 93.20% F1-Score for COVID-19 prediction.
• Experimental results confirmed the robustness and satisfactory performance of the deep learning model for COVID-19 detection.

Conclusions:

• Deep learning analysis of ECG data can effectively detect cardiovascular changes associated with COVID-19.
• ECG data holds potential as a non-invasive tool for the diagnosis of COVID-19.
• The developed framework can serve as a valuable decision-making support system for healthcare professionals in diagnosing COVID-19.