Deep multi-view feature learning for detecting COVID-19 based on chest X-ray images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Disease Diagnostics

Background:

• COVID-19 is a highly contagious pandemic disease requiring rapid and accurate detection.
• Chest X-ray imaging reveals abnormalities indicative of COVID-19, making it a valuable diagnostic tool.
• Early detection is crucial for controlling the transition of COVID-19.

Purpose of the Study:

• To develop and evaluate a multi-view feature learning method for detecting COVID-19 from chest X-ray images.
• To leverage deep learning models for accurate classification of COVID-19, normal, and pneumonia cases.
• To enhance the early diagnosis of COVID-19 by improving detection accuracy.

Main Methods:

• A multi-view feature learning framework was employed, integrating correlative and complementary information from deep features.
• Deep features were extracted using pre-trained Convolutional Neural Network (CNN) models: AlexNet, GoogleNet, ResNet50, SqueezeNet, and VGG19.
• Extreme Learning Machine (ELM) was utilized for the classification of extracted features.

Main Results:

• The proposed method achieved an overall accuracy of 99.82% for detecting COVID-19, normal, and pneumonia.
• Specific class accuracies were 100% (COVID-19), 99.82% (normal), and 99.82% (pneumonia).
• High sensitivity, specificity, precision, and F-scores were reported for all three classes, indicating robust performance.

Conclusions:

• The developed multi-view feature learning method demonstrates high effectiveness in detecting COVID-19 from chest X-rays.
• The findings suggest this method can serve as a valuable tool for expert assistance in early COVID-19 diagnosis.
• Accurate and early detection facilitates better management and control of the COVID-19 pandemic.