Location classification of lung nodules with optimized graph construction

The locations of lung nodules relative to the other lung anatomical structures are important hints of malignant cancers. In this paper, we propose a fully automatic method to identify if a lung nodule is well-circumscribed, juxta-vascular, juxta-pleural or pleural tail in computed tomography (CT) images. First, we design an optimized graph model, introducing new global and region-based energy terms, to label each voxel as background or foreground in a single graph cut algorithm. Then, the texture features of a lung nodule are extracted based on the voxel labeling outputs, and its location information is inferred. We evaluate the proposed method on low-dose CT images, and demonstrate highly effective nodule classification results comparatively.