An Optimal Model Selection for COVID 19 Disease Classification

Abstract

In the current scenario, the pandemic created by coronavirus is on the boom, and that is why it becomes very critical to control and cure this disease. The currently available technique for coronavirus disease 2019 (COVID-19) testing, i.e., reverse transcription polymerase chain reaction (RT-PCR), turns out to be a lot of time-consuming and requires modern labs, equipment, and highly trained medical staff that are rare to get. Chest computerized tomography (CT) is however available with a lot of ease, and it will be fruitful if these machines are used for COVID-19 testing. During this pandemic, there is an absolute need for an efficient and readily accessible way for COVID-19 patients classification, and CT is one of the best ways to do so. Keeping that in fact, this chapter introduces a study for understanding which deep learning models give the best result when classifying COVID-19 patients using chest CT images. For this study, ResNet 50, ResNet 101, DenseNet 121, DenseNet 169, and DenseNet 201 are compared with each other on the basis of classification accuracy, and it has been observed that DenseNet 169 has the tendency to yield best results with the accuracy of 96%. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.