COVID-19 Disease Identification on Chest-CT images using CNN and VGG16

TL;DR

This paper develops CNN and VGG16 models to automatically identify COVID-19 from chest CT images, achieving over 96% accuracy, providing a rapid diagnostic alternative to traditional testing methods.

Contribution

Introduces a deep learning approach using CNN and VGG16 architectures for COVID-19 detection on chest CT images, demonstrating high accuracy with a large public dataset.

Findings

CNN achieved 96.34% accuracy

VGG16 achieved 96.99% accuracy

Models provide rapid COVID-19 diagnosis from CT images

Abstract

A newly identified coronavirus disease called COVID-19 mainly affects the human respiratory system. COVID-19 is an infectious disease caused by a virus originating in Wuhan, China, in December 2019. Early diagnosis is the primary challenge of health care providers. In the earlier stage, medical organizations were dazzled because there were no proper health aids or medicine to detect a COVID-19. A new diagnostic tool RT-PCR (Reverse Transcription Polymerase Chain Reaction), was introduced. It collects swab specimens from the patient's nose or throat, where the COVID-19 virus gathers. This method has some limitations related to accuracy and testing time. Medical experts suggest an alternative approach called CT (Computed Tomography) that can quickly diagnose the infected lung areas and identify the COVID-19 in an earlier stage. Using chest CT images, computer researchers developed several deep learning models identifying the COVID-19 disease. This study presents a Convolutional Neural Network (CNN) and VGG16-based model for automated COVID-19 identification on chest CT images. The experimental results using a public dataset of 14320 CT images showed a classification accuracy of 96.34% and 96.99% for CNN and VGG16, respectively.