Deep Learning for Accurate Histopathological Classification of Lung and Colon Cancer Using CNN and EfficientNetB3 Models

This research work suggests a quick way to sort histopathological pictures of lung and colon cancers by using two deep learning models: a custom CNN and EfficientNetB3. In this respect, the proposed models were trained using a large dataset of both cancerous and benign tissues and tested for accuracy, precision, recall, and F1-score. The CNN model showed an overall accuracy of $\mathbf{9 9 \%}$, demonstrating its capability in differentiating between various types of cancers and benign conditions. EfficientNetB3 further optimized the performance, yielding $\mathbf{9 8 \%}$ accuracy with well-balanced metrics for all the classes with a high score in recognizing lung and colon adenocarcinoma and benign tissues. Robustness and good generalization on unseen data are further indicative of their potential clinical diagnostic utility. Steps for the future: rank pruning of slight misclassifications and application extension to other cancers than those under consideration by incorporating additional data and using advanced techniques, including transfer learning for real-time diagnostic support and detection at earlier cancer stages in healthcare.