In rural areas with limited access to diagnostic resources and treatment options, lung cancer poses a significant health risk.  More accessible decision-support tools are needed to assist with timely diagnosis and treatment planning.  Bayesian networks provide a framework for incorporating uncertainty and modeling relationships among patient characteristics, symptoms, risk factors, disease characteristics, treatments, and outcomes.  We will examine large lung cancer datasets and the preprocessing required to make them suitable for Bayesian network modeling before exploring methods for generating such structures.  We will focus primarily on score-based, constraint-based, and hybrid structure-learning methods and compare the resulting models in terms of their predictive performance and learned relationships.  Finally, we will compare our results with those of a previously published Bayesian network study of lung cancer survival and treatment selection to evaluate the potential of these approaches for clinical decision support.