Introduction of advantages of fluid dynamics simulations into clinical practice depends on ability to quickly create patient-specific computational models from computed tomography or magnetic resonance images with minimal human interaction. Our research addresses three steps in this process. First is identifying lung airways in radiological images. Our approach is based on a combination of geometric analysis of gradient vector flow and graph-based algorithms. Second is a novel method for robust and efficient computation of medial curves for complex biomedical geometries. This algorithm is based on three key concepts: a local orthogonal decomposition of geometry into substructures, a differential concept called the interior center of curvature, and integrated stability and consistency tests. Finally, we introduce variational method for generating prismatic boundary-layer meshes. Compared to existing methods, our approach is novel in the following aspects: it relies on feature size to make resulting mesh scale invariant and prevent global self-intersections; it uses face-offsetting method to propagate surface mesh to generate the high-quality prismatic layers; finally, it allows to add prismatic boundary layer to any tetrahedral mesh while preserving structural qualities of original discretization.