PNEXAI: An Explainable AI Driven Decipherable Pneumonia Classification System Leveraging Ensemble Neural Network

With an estimated two million deaths each year, pneumonia is a terrifying disease with a high mortality rate among children and the elderly. Due to a lack of medical surveillance in Africa and Asia, the poorest individuals are most susceptible to pneumonia. It is responsible for at least 28 percent of all child deaths in Bangladesh each year, and potentially much more. A number of computer-assisted diagnostic approaches for detecting pneumonia have been developed in recent years. This paper proposes the PNEXAI model to identify pneumonia using Chest X-Ray images using VGG16, VGG19, ResNet50, ResNet101, and Inception v3 models. At first, we collected data, categorized it, and trained the models. VGG16 achieved an accuracy rate of 97.17%, followed by VGG19 at 97.69%, ResNet50 at 97.35%, ResNet101 at 95.63%, and Inception V3 at 94.86%. The ensemble model comprised of the best three classifiers (VGG16, VGG19, and ResNet50) achieved the highest overall accuracy of 98.46%. Finally, in order to better understand our classification, we incorporated LIME driven XAI to our model.