Pneumonia Detection Using Ensemble Models

Pneumonia remains a leading cause of mortality globally, often challenging to diagnose accurately from chest X-rays due to its similarity to other lung conditions. This study introduces an automated pneumonia detection system leveraging an ensemble of Convolutional Neural Networks (CNNs)—DenseNet121, EfficientNetB0, and ResNet50—combined through a voting classifier. By harnessing the complementary strengths of these models, we achieve enhanced classification accuracy, robustness, and generalization compared to single-model approaches. The system includes a Gradio-based web interface for real-time predictions, making it accessible to healthcare professionals. Evaluated on datasets like Chest X-ray14, it achieves accuracy between 85-95%, with precision, recall, and F1-scores reflecting a balanced performance. Future enhancements, such as weighted ensembles and multi-disease detection, promise further improvements. This tool offers a practical solution for early pneumonia diagnosis, particularly in resource-limited settings. Key Words: Pneumonia Detection, Ensemble Learning, Convolutional Neural Networks, Chest X-Ray, DenseNet, EfficientNet, ResNet, Gradio Interface, Medical Imaging