Pneumonia Detection using Ensemble Transfer Learning

Pneumonia is among the most common illnesses and causes to death among the young children worldwide. It is more serious in under-developed countries as it is hard to diagnose due to the absence of specialists. Chest X-ray images have essentially been utilized in the diagnosis of this disease. Examining chest X-rays is a difficult task, even for an experienced radiologist. Information Technology, especially Artificial Intelligence, have started contributing to accurate diagnosis of pneumonia from chest X-ray images. In this work, we used deep learning, transfer learning, and ensemble voting to increase the accuracy of pneumonia detection. The models utilized are VGG16, MobileNetV2, and InceptionV3, all pre-trained on ImageNet, and used the Kaggle RSNA CXR image dataset. The results from these models are ensembled using the weighted average ensemble approach to achieve better accuracy and obtained 98.63% test accuracy. The results are promising, and the proposed model can assist doctors in detecting pneumonia quickly and accurately from Chest X-Ray.