Novel Feature Selection and Voting Classifier Algorithms for COVID-19 Classification in CT Images

Abstract

This research investigates the application of novel feature selection techniques and ensemble voting classifiers for improved COVID-19 detection from chest CT scan images. The primary objective is to develop a robust and accurate diagnostic tool that surpasses existing methods in terms of sensitivity and specificity. We explore a hybrid feature selection approach combining filter and wrapper methods, followed by a voting classifier that integrates multiple optimized machine learning models. Preliminary results indicate improved diagnostic performance compared to state-of-the-art techniques, offering potential for aiding clinicians in rapid and reliable COVID-19 diagnosis. This work contributes to improving the efficiency and accuracy of COVID-19 diagnosis from CT scans, potentially alleviating diagnostic bottlenecks during outbreaks.

Introduction

The COVID-19 pandemic highlighted the urgent need for rapid and accurate diagnostic tools. Chest CT scans provide valuable information for detecting COVID-19-related pneumonia, but manual interpretation is time-consuming and prone to inter-observer variability. Automated diagnostic systems using machine learning offer a promising solution. However, existing methods often face challenges related to high dimensionality of image data, class imbalance, and the need for improved diagnostic accuracy and robustness. This research addresses these challenges by developing a novel approach for COVID-19 detection from CT scans.

Objectives

• Develop a novel hybrid feature selection method to reduce dimensionality and improve classification accuracy.
• Design and implement a robust voting classifier that integrates multiple optimized machine learning models for improved COVID-19 detection.
• Evaluate the performance of the proposed system against existing state-of-the-art methods using relevant metrics (sensitivity, specificity, accuracy, AUC).