AI enabled pneumonia detection and diagnosis based on the concatenation approach: A framework for healthcare sustainability

Early detection and diagnosis of pneumonia play a significant role in saving human life. However, detection of pneumonia from chest X-ray images with the help of radiologists is a time-consuming task. Thus, the development of an appropriate artificial intelligence (AI) enabled model for the precise detection of pneumonia becomes an important research topic. In this aspect, we develop an automated transfer learning-based pneumonia detection framework using a feature concatenation approach. The proposed approach uses the DenseNet pre-trained network and concatenates the features extracted from several dense blocks of DenseNet in order to obtain the dense multiscale information from the chest X-ray images. This feature concatenation process helps in improving the classification accuracy of the proposed framework and simplifies the pneumonia detection process. The proposed work achieves accuracy, sensitivity, specificity, and precision of 98.60%, 97.03%, 99.14%, and 97.51%, respectively, on the chest X-ray pneumonia dataset which are superior results to the existing deep learning-based pneumonia frameworks. It is concluded that the proposed AI-enabled pneumonia detection framework has the prospective to be considered as a computer-aided diagnosis support system for the early diagnosis of pneumonia.