Classification of COVID-19 from chest x-ray images using deep features and correlation coefficient

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• COVID-19 pandemic necessitates effective diagnostic tools.
• Chest X-rays are crucial for identifying viral pneumonia, including COVID-19.
• Computer-aided diagnostic (CAD) systems can assist radiologists in early detection.

Purpose of the Study:

• To develop an optimized feature selection method for COVID-19 detection using chest X-rays.
• To classify chest X-ray images into COVID-19, Pneumonia, and Normal categories.
• To evaluate the performance of a proposed model against existing techniques.

Main Methods:

• Feature extraction using deep learning architectures (ResNet152, GoogLeNet).
• Feature space reduction via Pearson Correlation Coefficient (PCC) and variance thresholding.
• Multi-class classification using machine learning predictive classifiers.

Main Results:

• The proposed model demonstrated superior performance compared to previous related works.
• Effective feature reduction was achieved, improving classification accuracy.
• Validation on a dataset of 768 COVID-19 images and 5216 Pneumonia/Normal images.

Conclusions:

• The proposed PCC and variance thresholding method offers a promising approach for COVID-19 screening.
• Optimized feature selection enhances the reliability of CAD systems for respiratory disease diagnosis.
• Further analysis on COVID-19 specific image features could yield even more robust classification.