EVAE-Net: An Ensemble Variational Autoencoder Deep Learning Network for COVID-19 Classification Based on Chest X-ray...

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Machine Learning

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Traditional radiological methods like CT scans and X-rays for COVID-19 detection are time-consuming.
• Machine learning offers potential for improving diagnostic efficiency and accuracy.

Purpose of the Study:

• To propose and evaluate three EVAE-Net models for COVID-19 detection using chest X-ray images.
• To leverage variational autoencoders and ensemble techniques for enhanced feature extraction and classification.
• To provide an accurate and efficient automated method for identifying COVID-19.

Main Methods:

• Two encoders were trained on the COVID-19 Radiography Dataset to generate feature maps.
• Feature maps were concatenated and processed through reparameterization to create latent embeddings.
• Ensemble techniques were combined with variational autoencoders (EVAE-Net) for classification.

Main Results:

• The proposed EVAE-Net models demonstrated strong performance in detecting COVID-19.
• The best performing model achieved 99.19% accuracy for four-class classification.
• A high accuracy of 98.66% was achieved for three-class classification.

Conclusions:

• EVAE-Net models show significant potential for accurate and efficient COVID-19 detection.
• The integration of variational autoencoders and ensemble methods offers a promising approach for medical image analysis.
• This method could aid in faster disease diagnosis and transmission control.