Cancer is one of the leading causes of death in the world. Among all cancers, lung cancer is the number one cause of death. Early diagnosis of lung nodules can aid in reducing deaths related to lung cancer. In this chapter, we first provide an overview of conventionally used different feature extraction and selection approaches. We then discuss various feature learning and classification schemes, including strategies based on deep learning. We analyze a combination of deep features and hand-crafted features for the diagnosis of lung nodules. End-to-end training of a two-dimensional (2D) deep neural network is studied for improved diagnosis followed by an extensive analysis of an approach based on 3D CNN that incorporates high-level lung nodule attributes to improve diagnostic decision making. Finally, a discussion of current limitations and future trends related to deep learning for lung nodule diagnosis is presented.