Enhanced pneumonia prognosis via a hybrid deep learning ensemble: Dense Net, Efficient Net, and VGG16 integration

Pneumonia, a prevalent and potentially life-threatening lung infection, requires rapid and precise diagnosis for effective treatment. Conventional clinical methods are often time-consuming and may lack reliability. In this study, we propose a novel deep learning ensemble framework that integrates Dense Net, Efficient Net, and VGG16 architectures with a hybrid feature extraction approach combining Deep CNN and InceptionV3. To further enhance performance, a hybrid optimization strategy employing Bayesian Optimization (BO) and Particle Swarm Optimization (PSO) is applied for hyperparameter tuning. The proposed model demonstrates good performance, achieving 99.23% accuracy, 97.8% sensitivity, and 98.3% specificity in pneumonia detection, surpassing existing approaches. These results underscore the potential of deep learning to deliver a robust, clinically viable, and highly reliable solution for early and accurate pneumonia diagnosis.