Towards Robust Pneumonia Detection: An Ensemble Model Approach

Pneumonia is a respiratory illness that is particularly dangerous to children under the age of five, accounting for a high percentage of deaths in this group. With the advent of new technologies such as deep learning, researchers are optimistic about the potential to utilize these techniques for accurately distinguishing between normal and pneumonia-afflicted X-ray images. This paper tackles the problem of pneumonia detection by using a deep learning approach that incorporates Transfer Learning and an ensemble model, achieving an impressive accuracy rate of 96.31%. The methodology leverages three primary models—MobileNet, MobileNetV2, and DenseNet169. The resulting model surpasses the performance of several established models, effectively striking a balance in accurately discerning between normal lung images and those indicative of infection. The recommended method has the potential to improve pneumonia diagnosis accuracy while saving clinicians a significant amount of time.