Pneumonia Classification through Stacked Ensemble Learning and Fusion of VGG16 and VGG19 Architectures

Pneumonia is a type of infection that affects the lungs and, in some cases, might be life-threatening. Chest X-ray images can provide useful information for the diagnosis and treatment of pneumonia. Hence, building a classification model that can classify a given chest X-ray image can be of great use in the medical field. This paper aims to enhance the precision of pneumonia classification by utilizing convolutional neural networks. Various architectures like ResNet, GoogLeNet, VGGNet, and AlexNet have been evaluated. In doing so, their efficacy in feature extraction and image classification has been inspected to provide improvements in accuracy. Ensemble learning methods were further implemented in building the classification model for optimizing the model. The use of hyperparameter fine-tuning in this research and also the nuanced impact of learning rates, batch sizes, and other critical parameters to observe what effect they have on the accuracy of the classification model, thus contributing to a broader understanding of CNNs in medical analysis. After a thorough assessment of convolutional neural network architectures, the results indicate notable accuracies: GoogLeNet achieved 90.6%, AlexNet reached 85.73%, VGG16 attained 87.5%, and ResNet50 demonstrated an accuracy of 94.40%. Additionally, two ensemble models were introduced, with the fusion of VGG16 and VGG19 achieving a remarkable accuracy of 95.28%, and a stacked ensemble model with a noteworthy accuracy of 95.75%.