USING CONVOLUTIONAL NEURAL NETWORK (CNN) FOR COVID-19 CHEST X-RAY DIAGNOSIS

Abstract

Artificial intelligence (AI) was used in chest X-ray (CXR) picture data categorization to brush on accuracy in diagnosing and promptly detecting pneumonia caused by COVID-19. The purpose of this study is to center on developing computer vision which basis for programs for the discovery of radiological features related to COVID-19 on the radiological pictures of the lungs and helps doctors provide appropriate treatment measures to reduce mortality. Over time, CNNs have consistently outperformed other image classification algorithms. Therefore, three architectures based on CNNs including VGG16, ResNet50, and ResNet101 were used to support the successful diagnosis of COVID-19 pneumonia on CXR images. Among them, the best classification performance is the ResNet101 model in terms of average Accuracy and Loss function which are obtained as 95.42% and 0.1492, respectively. Additionally, ResNet101 performed well in classifying the COVID-19 cases in the test dataset with a Precision, Recall, and F1-score of 92.2%, 94.0%, and 93.0%, respectively. Hence, the ResNet101 model can help in clinical practice by building a website that facilitates health professionals to save time and reduce manual errors during the diagnosis.