Deep Learning-Based Lung Nodule Classification and Lung Cancer Diagnosis: A Comprehensive Literature Review

Abstract

Lung cancer is still one of the major causes of cancer-related deaths in the world and timely and precise diagnosis is very important to enhance survival of the patients. The past decade has seen remarkable progress in the medical image analysis arena for detection, segmentation and classification of pulmonary nodules, which is driven by the recent developments in Artificial Intelligence (AI) and Deep Learning (DL). This literature review aims to provide a thorough overview of recent research articles published from 2025 to 2026 related to lung nodule classification based on computed tomography (CT) and chest radiography images within the context of deep learning. The review delves into the mechanisms of several architectures such as Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), ensemble learning, transfer learning, multi-task learning, Neural Architecture Search (NAS), and Explainable AI (XAI). Furthermore, the segmentation methods and data augmentation strategies are explored, as well as being evaluated in conjunction with hybrid AI frameworks. The review points out the advantages, weaknesses, and the data sets, measures, and clinical utility of the available approaches. Results suggest that transformer-based models, hybrid CNN-transformer models, and explainable ensemble models outperform the classification accuracy of other models, which is above 95%, in several research studies. There are many challenges that are yet to be solved, including the limited annotated datasets, interpretability of the models, generalization across clinical environments and computational complexity. Multimodal learning, federated learning, integrating explainable AI, and clinically validated real-time diagnostics are directions for future research.