Lung and colon cancers are among the most prevalent and deadly cancer types worldwide, with early detection being crucial for improving survival rates. Traditional diagnostic methods, such as manual histopathological image analysis, are time-consuming, prone to errors, and often inconsistent. To address these challenges, this study proposes a dual-branch deep learning framework combining ResNet and EfficientNet architectures for the accurate detection of lung and colon cancers from histopathological images. By leveraging the strengths of both ResNet’s residual learning and EfficientNet’s efficient scaling, the framework aims to improve the classification performance while mitigating issues like overfitting, poor generalization, and class imbalance. The study also explores the potential of a multi-branch approach by integrating ResNet, Xception, and DenseNet, further enhancing model robustness and accuracy. Comprehensive experiments are conducted on a large dataset of cancerous histopathological images, demonstrating the superior performance of the proposed models in terms of accuracy, precision, recall, and F1-score. The results show that the dual and multi-branch models significantly outperform existing approaches, highlighting their potential for real-world clinical applications in cancer diagnosis.

Keywords: Lung Cancer, Colon Cancer, Deep Learning, ResNet, EfficientNet, Xception, DenseNet, Multi-Branch Models, Dual-Branch Models, Histopathological Image Analysis, Overfitting, Class Imbalance, Model Generalization, Cancer Detection.