Enhanced COVID-19 Detection in CT Images using Preprocessed Dense Convolutional Neural Networks

Abstract

This study addresses the need for rapid and accurate COVID-19 diagnosis by utilizing Convolutional Neural Networks (CNNs) on CT chest images. Given the limitations of traditional RT-PCR testing, CT imaging has emerged as a highly sensitive diagnostic tool, but manual analysis can be time-consuming and error- prone. This research investigates the impact of data augmentation and median filtering on COVID-19 classification performance using CT images. Experiments were conducted with two pre-trained CNN models DenseNet121 and DenseNet169 across three conditions: (1) COVID with data augmentation, (2) COVID without data augmentation, and (3) COVID with median filter and data augmentation