Lung cancer is one of the leading causes of cancer-related deaths worldwide, making early and accurate detection essential. This study presents a CNN-based framework using the pre-trained VGG16 model for automated lung cancer classification from chest CT scans. The approach performs both binary (cancerous vs. non-cancerous) and multi-class classification of four subtypes: Adenocarcinoma, Large Cell Carcinoma, Squamous Cell Carcinoma, and Normal tissue. Transfer learning and data augmentation were employed to mitigate limited data and improve generalization, while focal loss enhanced recognition of minority classes. The proposed model achieved 99.68% accuracy for binary and 89.21% for multi-class classification, outperforming benchmark models. These results demonstrate that the method is robust, reliable, and clinically relevant. Future work will focus on external validation and interpretability to strengthen its potential as a computer-aided diagnostic tool.
