Lung cancer (LC) is one of the most lethal malignancies worldwide, and early detection is essential. This study develops a deep learning (DL) based classification model for LC diagnosis using computed tomography (CT) images. In the experiments conducted on the IQ-OTHNCCD LC dataset, the Synthetic Minority Over-sampling Technique (SMOTE) method was applied to eliminate class imbalance, data augmentation techniques were used, and an early stopping mechanism was integrated to enhance the model's generalizability. Commonly used convolutional neural network (CNN) architectures, such as ResNet101, VGG19, and DenseNet121, are compared, and the model's performance is analyzed in detail. With an accuracy of 98%, the trial results demonstrate that the suggested ResNet101 model offers the best classification performance. the DenseNet121 model exhibited a relatively lower accuracy rate in distinguishing between benign and normal classes. The study conclusively demonstrates that an optimized ResNet101-based deep learning model, enhanced with data balancing and augmentation techniques, provides the most accurate and reliable classification performance for lung cancer detection using CT images. It not only outperforms traditional CNN architectures in terms of overall accuracy (98%) but also achieves perfect classification in malignant cases. These results validate the model’s potential as a robust diagnostic aid and highlight its superiority over existing methods in the literature, particularly in handling class imbalance and maintaining generalization across diverse image types.