Lung cancer is among the most prevalently diagnosed and fatal cancers globally; thus, it emphasizes the need for an efficient detection system that can enhance diagnostic performance and patient care. The work reviews recent research in the area of lung cancer detection through deep learning, primarily convolutional neural networks (CNNs) and hybrid models intended to process CT scan images. A review of approaches on popular datasets like LIDC-IDRI, LUNA16, NLST, and some clinical datasets is comparing the performances of various methodologies like ensemble 2D CNNs, GoogleNet-based CNNs, and multi-models boosted by optimization methods. CNN-based approaches perform at an accuracy rating (91–97%) through various datasets which is supported by many performance metrics including accuracy, precision, recall, and F1-scores. Additional studies introduced specific optimization techniques, such as Ebola Optimization and Taguchi Parametric Optimization, to improve these models’ reliability and robustness. Overall, these results demonstrate the power of deep learning for lung cancer detection and suggest that it could serve as a valuable tool in clinical decision-making and aid in early diagnosis. Yet challenges remain such as dataset variability and cross-platform interoperability. Future research will target the improvement of models to achieve better generalizability and synthesis of clinical knowledge into the AI diagnostic systems.