Biphasic majority voting-based comparative COVID-19 diagnosis using chest X-ray images

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Diagnostic Systems

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Expert radiologists are limited, increasing the need for automated diagnostic methods.
• Chest X-rays are valuable for diagnosing COVID-19, pneumonia, and other conditions.

Purpose of the Study:

• To develop and evaluate an automated system for diagnosing COVID-19, pneumonia, and normal cases using chest X-ray images.
• To compare the performance of multiple classifiers in disease diagnosis.
• To enhance the reliability and usability of AI-driven diagnostic tools.

Main Methods:

• Utilized a dataset of chest X-ray images for COVID-19 Positive/Negative and Normal/Pneumonia classification.
• Employed a biphasic majority voting method combining five successful classifiers (KNN, Linear Discriminant, Logistic Regression, Bagged Trees Ensemble, SVM).
• Extracted image features using the Bag of Features method.

Main Results:

• Achieved high accuracy rates: 99.86% in Phase-1 and 99.28% in Phase-2.
• The overall system accuracy reached 99.63%.
• Demonstrated superior performance metrics (Specificity, Precision, Recall, F1 Score, AUC, MCC) compared to existing literature.

Conclusions:

• The proposed automated system demonstrates exceptional accuracy and reliability for diagnosing COVID-19, pneumonia, and normal conditions from chest X-rays.
• The biphasic majority voting technique enhances the robustness of the diagnostic model.
• The study highlights the importance of usability alongside performance in AI healthcare applications.