💻 computer science

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

RespGeomLib is a reproducible parametric engine that generates analysis-ready, seamless human airway lumen geometries from compact specifications by combining port-based assembly with implicit smooth-min blending, offering superior junction quality and computational efficiency compared to existing methods for pulmonary morphometry and CFD simulations.

Original authors: Nichula Wasalathilaka, Parakrama Ekanayake, Roshan Godaliyadda

Published 2026-06-19

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Nichula Wasalathilaka, Parakrama Ekanayake, Roshan Godaliyadda

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine trying to build a perfect model of the human breathing system (the airways) to study how air flows through it. Usually, scientists try to do this by taking CT scans of real people and trying to trace the tubes. But this is like trying to trace a winding river on a blurry, low-resolution map: the ends of the river often disappear, and the places where rivers split (bifurcations) look jagged and messy. Cleaning up these digital models is a nightmare, and the messy joints can ruin computer simulations.

RespGeomLib is a new tool that solves this by building the airways from scratch using a "recipe" instead of a blurry map. Here is how it works, explained simply:

1. The "Lego" Approach (Parametric Engine)

Instead of trying to copy a messy real-world scan, RespGeomLib uses a YAML file. Think of this as a simple, human-readable shopping list or recipe. You tell the computer: "Make a tube 5cm long, 2cm wide, then split it into two smaller tubes at this specific angle."

The Magic: Because it's built from a recipe, you can recreate the exact same airway model over and over again. If you need to test a new idea, you just tweak the numbers in the recipe, and the computer rebuilds the model instantly.

2. The "Smoothie" Junctions (Implicit Blending)

The biggest problem with old methods is how they connect tubes. Imagine gluing two plastic straws together; you usually get a sharp, ugly seam or a weird lump where they meet.

The Old Way: It's like stitching two pieces of fabric together. You can see the needle holes and the seam is stiff.
The RespGeomLib Way: It uses a mathematical "smoothie" technique. Instead of gluing the tubes, it treats them like soft clay or liquid. Where the tubes meet, the computer blends them together so the surface flows seamlessly, like two streams of water merging into one. There are no sharp edges, no seams, and no "internal walls" blocking the air.

3. The "Smart Builder" (Hybrid Strategy)

Building a whole tree of airways using this "clay" method for every single tiny branch would take forever and crash your computer (like trying to sculpt a whole forest out of clay one leaf at a time).

The Solution: RespGeomLib is smart. For the long, straight parts of the tubes, it uses fast, simple math (like drawing a straight line). It only switches to the heavy-duty "clay sculpting" math right at the junctions where tubes split. This makes it incredibly fast and efficient, saving huge amounts of computer memory.

4. What Can You Do With It?

The paper shows three main things this tool can do:

Cleaner Models: It creates junctions that are mathematically perfect and "watertight" (no leaks), which is essential for accurate computer simulations.
Controlled "What-If" Scenarios: You can take a perfect model and easily create a "sick" version (narrowed tubes) or a "stretched" version (wider tubes) without breaking the model. This is great for testing how different blockages affect airflow.
Ready for Simulation: The final model comes out with clean, open ends (ports) ready to be plugged into airflow software. The authors tested this by running a wind-tunnel-style simulation (CFD) and showed that the air flowed smoothly and the computer didn't crash.

What It Is NOT

It is important to note what this paper doesn't claim:

It is not a tool for diagnosing a specific patient's disease right now. It doesn't look at a real person's CT scan and fix it.
It is not a replacement for real anatomy. It creates "idealized" or "synthetic" airways based on rules, not unique biological quirks like a specific person's crooked nose or uneven lung shape.

In a nutshell: RespGeomLib is a digital factory that builds perfect, reproducible, and smooth breathing tube models from simple text instructions. It fixes the "ugly seams" of old methods and does it so fast that scientists can now run complex airflow simulations without their computers exploding.

Technical Summary: RespGeomLib

Problem Statement
Current methods for generating human airway geometries face significant limitations depending on the source. CT-derived models, while anatomically grounded, are constrained by distal scan resolution, sensitive to image noise and partial-volume effects, and often require extensive manual cleanup near bifurcations before they are suitable for morphometric analysis or Computational Fluid Dynamics (CFD). Conversely, procedural alternatives offer reproducibility but frequently rely on stitching tubular primitives. This approach introduces non-smooth junctions, sharp seams, and poorly defined open boundaries that degrade downstream measurement and simulation quality. Furthermore, applying implicit surface modeling to an entire airway tree is often computationally prohibitive due to the large spatial extent and significant scale variations from proximal to distal branches.

Methodology
The authors present RespGeomLib, a reproducible parametric engine that generates analysis-ready human airway lumen surfaces from compact, human-readable YAML specifications. The framework employs a hybrid extraction strategy to balance geometric flexibility with computational efficiency:

Parametric Primitives and Assembly: The system utilizes three geometric primitives: Pipe (tapered tubes), Y2 (two-way junctions), and Y3 (three-way junctions). These are defined in local canonical frames with explicit inlet/outlet ports. Assembly is achieved via a port-based interface where child primitives are attached to parent ports using rigid transforms (rotation and translation) defined by elevation–azimuth angles.
Smooth Junction Blending: To avoid the seams associated with stitched primitives, junctions are modeled using an implicit smooth-min formulation. The geometry is defined as the zero level-set of a blended scalar field:
$\phi(x) = -\frac{1}{\kappa} \log \left( \sum_{i=1}^{N} e^{-\kappa \phi_i(x)} \right)$
where $\phi_i(x)$ represents the distance to the $i$ -th branch centerline and $\kappa$ controls the blend sharpness. This ensures seamless, lumen-continuous junctions without internal walls.
Local-Implicit Extraction: To avoid the cost of full-tree voxelization, the engine applies implicit isosurface extraction (via marching cubes) only locally on tight grids around junctions. Straight and tapered segments are generated analytically. The local grid spacing is adaptive, defined relative to the local branch radii.
Controlled Variants and Growth: The framework supports morphometry-guided tree generation (e.g., Weibel/ICRP models) and the creation of synthetic variants. Localized lumen perturbations (stenosis or dilation) are applied via a parametric radius modulation function $r(s) = r_0(s)(1 + \alpha w(s))$ , preserving the underlying topology and port definitions.
Post-Processing: The pipeline enforces open-port boundaries by clipping surfaces at port planes, followed by deterministic cleanup (removing duplicates/degenerates) and manifoldness validation to ensure CFD readiness.

Key Contributions

RespGeomLib Engine: A YAML-based parametric system that generates triangulated lumen meshes with explicit inlet/outlet ports from compact specifications.
Smooth Junction Construction: A smooth-min implicit formulation that produces seamless bifurcations and trifurcations, eliminating stitched seams and internal walls.
Efficient Hybrid Meshing: A local-implicit/analytic pipeline that avoids full-tree voxelization, demonstrated to be significantly faster and more memory-efficient than global implicit extraction.
Reusable Downstream Utility: Demonstrations of controlled synthetic variant generation, morphometry-guided tree synthesis, and successful export for CFD simulations.

Results

Junction Quality: Compared to a Boolean/stitch baseline, RespGeomLib produced junctions with 0 non-manifold edges (vs. 6 in the baseline) and near-perfect port planarity (maximum error 0.0010 vs. 5.9796). The baseline failed planar capping tests, whereas RespGeomLib successfully generated watertight surfaces.
Computational Scaling: In matched-depth comparisons of balanced Y2 trees, the local extraction strategy was 76.7× to 279.8× faster than global implicit extraction at depths 1–3. It also utilized 1.11× to 1.31× less peak memory. The global baseline became computationally expensive even at shallow depths.
CFD Usability: A steady incompressible RANS simulation on a RespGeomLib Y-junction demonstrated stable solver behavior. The case yielded physically plausible flow redistribution (asymmetric split of 46.5%/53.5%), a pressure drop of $\approx 1.0 \times 10^{-1}$ Pa, and final residuals on the order of $10^{-5}$ .
Synthetic Variants: The system successfully generated labeled cohorts of stenotic and dilated airways from a fixed base geometry, maintaining consistent topology and port definitions.

Significance and Claims
The paper positions RespGeomLib not as a patient-specific reconstruction tool, but as a reproducible geometry engine for generating analysis-ready lumen surfaces. Its primary significance lies in providing a compact, editable, and simulation-ready representation that bridges the gap between anatomical reconstruction and procedural modeling.

The authors claim that the hybrid approach (analytic segments + local implicit blending) offers a tractable solution to the computational costs of global implicit modeling while overcoming the geometric defects of stitched primitives. This enables:

Reproducible Morphometry: Consistent generation of airway trees based on standard morphometric rules.
Controlled Benchmarking: The ability to generate synthetic variants with specific, labeled geometric perturbations (e.g., stenosis) for sensitivity analysis.
Simulation Readiness: The production of watertight, manifold meshes with well-defined boundary ports suitable for immediate use in CFD workflows.

The study acknowledges limitations, noting that it focuses on idealized lumen geometry and does not yet model patient-specific asymmetry, non-circular cross-sections, or wall thickness. The CFD results are presented as a usability demonstration rather than a comprehensive fluid-dynamics benchmark.