An efficient deep learning approach for detecting lung disease from chest X-ray images using transfer learning and ensemble modeling

Abstract

Among the most convenient bacteriological assessments for the diagnosis and treatment with several health complications is the chest X-Ray. The World Health Organization (WHO) estimates, for instance, that pneumonic plague induces between 250,000 to 500,000 fatalities annually. Pneumonia and flu are serious challenges towards global health as well as being a source of significant death rates globally. [1]. In X-Ray imaging, it is a common technique to standardize the extracted image reconstruction with usual uniform disciplines taken before the study. Unfortunately, there has been relatively little study on several separate lung disease monitoring, including X-Ray picture analysis and poorly labelled repositories. Our paper suggests an effective approach for the detection of lung disease trained on automated chest X-ray images that could encourage radiologists in their moral choice. Besides, with a weighted binary classifier, a particular technique is also deployed that will optimally leverage the weighted predictions from optimal deep neural networks such as InceptionV3, VGG16 and ResNet50. In addition to the existing, transfer learning, along with more rigorous academic training and testing sets, is used to fine-tune deep neural networks to achieve higher internal processes. In comparison, 88.14 percent test accuracy was obtained with the final proposed weighted binary classifier, where other models give us about 76.91 percent average accuracy. For a brief recurring diagnosis, the legally prescribed procedure may also be used which may increase the course of the same condition for physicians. For a prompt diagnosis of pneumonia, the suggested approach should be used and can improve the diagnosis process for health practitioners.