Detection of Lung Infection on CT Scan for Covid-19 Disease Using Sparrow Search Based Deep Learning Model

Rapid globalization of the COVID-19 virus was observed at the start of 2018. The prevention and treatment of this illness are crucial. Imaging techniques such as chest computed tomography (CT) scans and RT-PCR can be used to categorize COVID-19 more accurately in the epicenter of the outbreak. Hospital reports have indicated that RT-PCR assays are not very sensitive when used to diagnose an infection in its early stages. This has led to calls for a diagnostic method that can quickly and accurately spot the Covid-19. CT has been proven to be an effective diagnostic tool. This study investigates the application of convolutional neural networks (CNNs) for the detection of COVID-19 in lung images. We propose a bi-channel CNN that combines gray-level entropy and pre-processed images using unsharp masking. The model was trained on a dataset of lung CT scans and evaluated for its accuracy in detecting COVID-19. The outcomes demonstrated that the suggested approach aided radiotherapists in making a speedy and exact analysis of COVID-19, achieving a prediction accuracy of 93.78%, and a false-negative rate of only 6.5%. These results indicate the potential of the bi-channel CNN to enhance diagnostic accuracy and efficiency in clinical settings. This novel approach addresses the limitations of traditional RT-PCR tests and manual CT scan analysis, offering a robust tool for early and accurate COVID-19 detection. For additional verification of the quality of the projected model, we used the SARS-COV-2-CT-Scan benchmark dataset. The outcomes demonstrated that the suggested approach can aid radiotherapists in making a speedy and accurate analysis of COVID-19.