An Advanced Deep Learning Framework DeepLungNet with Global Average Pooling for Precise Lung Cancer Classification

IntroductionLung cancer remains one of the deadliest diseases worldwide, with survival rates heavily dependent on early and precise detection. However, existing diagnostic methods face challenges in accuracy and efficiency. This research introduces DeepLungNet, an enhanced deep learning framework designed to improve lung cancer classification using Global Average Pooling (GAP). GAP is immensely important in reducing model complexity and overfitting, making it ideal for imaging in healthcare where datasets happen to be constrained. By preserving critical spatial information, it enhances model generalization and ensures reliable predictions. MethodsTo develop a robust classification model, chest X-ray images were gathered from varuious origins, ensuring high diversity and quality. The dataset includes five categories: Bacterial Pneumonia, COVID-19, Normal, Tuberculosis, and Viral Pneumonia. DeepLungNet was evaluated against standard CNN architectures, including CNN, ResNet-50, and VGG16. The GAP layer proved instrumental in improving training efficiency, reducing computational overhead, and boosting classification accuracy. This makes it a strong candidate for real-world clinical deployment, particularly in resource-limited settings. ResultsDeepLungNet achieved 100% accuracy, outperforming conventional models: CNN (93.40%), VGG16 (94.25%), and ResNet-50 (95.75%). The model also attained perfect precision, recall, and F1-score, reinforcing its reliability in lung disease detection.ConclusionDeepLungNet demonstrates exceptional performance, making it a viable solution for accurate and efficient lung disease classification. Its 100% accuracy and reduced computational demands make it ideal for clinical applications requiring fast, dependable diagnoses.