Publication

Abstract : The accurate detection of pneumonia using chest X-rays is crucial for rapid medical treatment. Despite advancement, achieving high accuracy and recall rates remains challenging due to the complexity of image processing and data imbalance. In this paper, we propose a novel ensemble model for pneumonia detection that utilizes concatenation, averaging, and weighted averaging techniques. Using pretrained models like VGG19, InceptionV3, and Vision Transformer, our method delivers impressive precision (0.9861), recall (0.9813), F1 score (0.9816), and accuracy (0.9899). Interestingly, the concatenation technique surpasses all metrics. We address the issue of data imbalance by merging two standard datasets, which improves the robustness of our technique. Comparative research demonstrates that our technology is superior than existing methods. Our results highlight the value of curated datasets in healthcare research and show how ensemble learning may improve the accuracy of pneumonia diagnosis.  © 2024 IEEE