An analysis of chest x-ray image classification and identification during COVID-19 based on deep learning models

Abstract

The COVID-19 epidemic profoundly affected both the global health and economic landscapes. The restriction of COVID-19 test kits is the reason for the inevitable delay in the diagnosing procedure. Therefore, the development of more economical and accessible diagnostic techniques is urgently needed. A chest X-ray is a crucial first step in obtaining a positive COVID-19 diagnosis since it makes it simple to identify any abnormalities in the chest. The project's overarching goal is to develop a picture classification algorithm that can accurately anticipate a presence of COVID-19 in CXR images by making use of pre-trained models. CNNs' recognition ease and capacity to classify images based on pertinent features make them ideal for use in medical image classification applications. This work used four DL-based classification algorithms-EfficientNetB1, VGG-16, COVID-Net, and ResNet-101-to categorise COVID-19 CXR pictures. The dataset utilised was the COVID-19 radiography dataset. Common classification metrics, like recall, accuracy, precision, and confusion matrices, are used to assess the efficacy of these algorithms. Results show that ResNet-101 achieves a best outcomes according to accuracy, precision, and recall (96%), indicating that it can be utilised for optimal COVID-19 diagnosis and offers a more cost-effective and time-saving alternative to other models.