Lung cancer remains a leading cause of cancer-related mortality worldwide, underscoring the urgent need for early and accurate diagnosis. Despite substantial progress in artificial intelligence (AI) and deep learning (DL), current diagnostic methodologies often encounter challenges related to data heterogeneity, model generalizability, and interpretability. This paper provides a detailed review of contemporary AI-based approaches for lung cancer diagnosis using chest CT scans, with a special focus on the role of texture-based features in improving classification performance. Identified gaps include the underutilization of hybrid architectures and the limited interpretability of deep learning models. In response, this study presents a novel hybrid framework combining Vision Transformers (ViT) and CNN-GRU architectures for enhanced spatial and temporal analysis of CT scan data. The proposed model has been fully implemented and evaluated using real-world CT scan data, achieving strong performance metrics. The paper concludes with suggestions for future research aimed at improving explainability, multimodal integration, and real-world clinical deployment.