An Efficient Pneumonia Detection from the Chest X-Ray Images

Currently, the human race is experiencing an unprecedented health crisis due to the outbreak of the COVID-19 pandemic. As per the World Health Organization, Pneumonia is one of the symptoms of severe stages of the infection. Pneumonia is itself a fatal respiratory infectious disease. X-Ray is the most reliable tool for the diagnosis of Pneumonia and its variants. It requires a human expert to conclude the observations from the obtained X-Ray images. It creates an overwhelming pressure on human resources in a pandemic. An automated solution could be an efficient and reliable alternative. In this paper, we have used deep learning approaches to diagnose Pneumonia and further classify them as viral or bacterial. A comparative study has been made to prove that deep learning-based pneumonia image classification is a valid argument for automating the diagnosis process through the web or mobile applications. VGG16, VGG19, ResNet50, MobileNetV1, and EfficientNetB3 are implemented and compared with one another based on the obtained accuracy, f-measure, and AUC metrics. The EfficientNet-B03 provides \(93\%\) and 88.78% accuracies for the binary and ternary experiments for chest X-Ray image classification, respectively. The said results are the reported best-performing accuracies for the used dataset, to the best of our knowledge.