Viral Pneumonia Detection Using Modified GoogleNet Through Lung X-rays

Abstract

Viral pneumonia occurs in lungs by the viral infection and is a life threating disease if not treated at the right time. Age is a critical factor in this regard and the effect of viral pneumonia varies from person to person diversly. People of an older age and infants are at considerable risk due to viral pneumonia that affects the lungs. It is difficult and time consuming for radiologists to detect the viral pneumonia by manually analyzing the lungs x-rays. So, Deep learning-based approaches are employed on lung x-rays for the accurate detection of viral pneumonia disease to avoid wrong medication. Therefore, it is necessary to propose an automated method that can accurately detect viral pneumonia patients to assist the medical doctors in their decision-making process. In this paper, we employed three different pretrained models such as AlexNet, GoogleNet, and ResNet18 to investigate the performance of transfer learning-based classification task to detect the viral pneumonia patients. We fined tuned all the three models. Along with this, we applied data augmentation to increase the amount of data to avoid the overfitting problem, which is common if the data is small for training the model. Among the three pretrained customized models, we achieved remarkable performance results on GoogleNet and obtained remarkable accuracy of 96.64%, precision of 96.99%, recall of 96.26%, specificity of 97.01%, and F1-score of 96.63%. More specifically, our method effectively detected the viral pneumonia patients accurately and precisely. Experimental results and comparative analysis with existing state-of-the-art methods illustrate that our method is useful in reliable detection of viral pneumonia patients and can be used by radiologists in their decision-making process.