RespGeomLib: A Reproducible Parametric Engine for Generating Analysis-Ready Human Airway Lumen Geometry

RespGeomLib is a parametric engine for building human airway lumen surfaces from compact YAML descriptions, aimed at workflows where reproducibility and simulation quality matter more than hand-authored modeling. The core idea is to assemble airway branches through ports and then blend junctions with an implicit smooth-min approach, so the resulting bifurcations are seamless instead of looking like stitched tubes.

For developers, the interesting part is not just the medical domain: it’s the geometry pipeline. The authors explicitly avoid full-tree voxelization by combining analytic segments with local implicit extraction only around bifurcations. That should reduce the memory and runtime costs that usually make global implicit methods awkward for large procedural trees.

They report cleaner junctions than a Boolean/stitch baseline, plus substantially better speed and memory efficiency than whole-tree global implicit extraction. The paper also mentions morphometry-guided tree generation, controlled synthetic airway variants, and CFD-ready export with stable airflow simulation, which suggests the system is meant for repeatable experiments rather than one-off meshes.

This is most relevant to graphics programmers and technical artists who build procedural geometry tools, but the broader lesson is useful too: if you need generated assets that survive downstream analysis or simulation, local implicit blending and compact declarative specs can be a strong alternative to brute-force mesh surgery. The code is publicly available, and the paper was submitted to arXiv on 17 Jun 2026.