A Lattice Boltzmann Method for Non-Newtonian Blood Flow in Coiled Intracranial Aneurysms

This paper presents a patient-specific Lattice Boltzmann method that models non-Newtonian blood flow in coiled intracranial aneurysms by treating the coil as an inhomogeneous porous medium, demonstrating the approach's validity for assessing post-treatment hemodynamics.

The Gist

The Big Picture: Fixing a Leaky Dam

Imagine a brain artery as a river. Sometimes, the riverbank gets weak and puffs out, forming a bubble called an aneurysm. If this bubble bursts, it causes a dangerous type of stroke.

To stop this, doctors often insert tiny metal springs (called coils) into the bubble. Think of these coils like stuffing a water balloon with steel wool. The goal is to slow down the rushing water inside the bubble so it stops hitting the weak walls, without blocking the main river flow.

The problem is: How do doctors know exactly how many springs to use and where to put them? It's currently a guessing game based on experience. This paper tries to build a "digital crystal ball" to predict exactly what happens to the blood flow once those springs are inserted.

The Challenge: Too Many Tiny Details

To simulate this on a computer, you usually have two choices:

1. The "Super-Resolution" Camera: You try to draw every single tiny wire of the metal spring in the computer model. This is incredibly accurate but takes a massive amount of computer power and time. It's like trying to count every single grain of sand on a beach to understand the tide.
2. The "Foggy Lens" Approach: Instead of drawing every wire, you treat the area with springs as a "sponge." You don't see the individual wires, but you know the sponge slows the water down. This is much faster but might miss some tiny details.

What This Paper Did

The authors created a new computer method that combines the best of both worlds. They used a technique called the Lattice Boltzmann Method (LBM).

• The Metaphor: Imagine the blood isn't a smooth river, but a crowd of millions of tiny people running. The LBM tracks how these "people" bump into each other and move.
• The Twist: Blood isn't like water; it's thick and sticky (like ketchup). The authors made their computer model smart enough to handle this "sticky" behavior (called non-Newtonian flow).
• The Innovation: They successfully taught this "crowd simulation" how to handle the "sponge" (the coil area) without needing to draw every single wire. They treated the coil area as a porous medium where the water has to squeeze through tiny holes.

The Experiment: Testing the "Digital Twin"

The team took a real patient's brain scan (anonymized) and built a digital 3D model of their aneurysm. Then, they ran three different simulations:

1. No Springs: Just the bubble.
2. Full Detail: They simulated the springs wire-by-wire (the "Super-Resolution" way).
3. The Sponge Model: They simulated the springs as a porous sponge (the "Foggy Lens" way).

They tested three different "packing densities" (how tightly the springs were stuffed): 15%, 20%, and 25%.

The Results: The Sponge Works!

The paper claims that the "Sponge Model" (Volume-Averaged approach) was almost identical to the "Full Detail" model.

• The Flow: In both models, the springs successfully slowed down the blood inside the bubble by about 50%. The water still flowed through the main river, but the dangerous swirling inside the bubble stopped.
• The Pressure: The force of the water hitting the bubble walls (Wall Shear Stress) dropped significantly in both models. This is good news because less force means a lower risk of the bubble bursting.
• The Speed: The "Sponge Model" was just as accurate for the big picture but much faster to compute.

The Conclusion

The authors say this workflow allows doctors to look at a specific patient's brain scan and simulate different treatment options before ever touching the patient.

They found that you don't need to simulate every single wire to get a good answer. Treating the coil area as a "sponge" is a valid, accurate, and much faster way to predict if a treatment will work.

What they didn't claim:

• They did not say this method is currently being used in hospitals to treat patients today.
• They did not claim this solves the problem of where to place the coils perfectly (they noted that coil placement varies and is still a challenge).
• They did not claim this eliminates the need for a doctor's experience, but rather offers a tool to help them.

In short: They built a fast, accurate computer game that predicts how blood behaves in a brain aneurysm after it's been stuffed with metal springs, proving that you can skip the tiny details and still get the right answer.

Technical Summary

Technical Summary: A Lattice Boltzmann Method for Non-Newtonian Blood Flow in Coiled Intracranial Aneurysms

Problem Statement
Intracranial aneurysms are a primary cause of hemorrhagic stroke, often treated via coil embolization, where metallic coils are inserted to induce thrombosis and reduce blood flow. However, predicting the hemodynamic outcomes of such treatments remains an open challenge. Accurate numerical prediction requires patient-specific geometry and a representation of the inhomogeneous coil distribution. Traditional approaches face a trade-off: fully resolved simulations of complex coil geometries are computationally expensive, while simplified models may fail to capture critical flow heterogeneity. Furthermore, blood is a non-Newtonian fluid, necessitating rheological models that account for shear-thinning behavior, which adds further complexity to the simulation.

Methodology
The authors propose a workflow combining patient-specific geometry with a volume-averaged approach to model the coiled aneurysm as an inhomogeneous porous medium. The core methodology involves:

1. Governing Equations: The fluid dynamics are described by the Volume-Averaged Navier–Stokes Equations (VANSE) for non-Newtonian flow. These equations incorporate porosity ($\phi$) to represent the fluid volume fraction and include interaction forces (Darcy and Forchheimer drag) to model the resistance caused by the coils.
2. Rheology: The non-Newtonian nature of blood is modeled using the Carreau–Yasuda viscosity model, which relates dynamic viscosity to the shear-strain rate.
3. Numerical Solver: The equations are solved using a problem-adapted Lattice Boltzmann Method (LBM) on a D3Q27 lattice. Key features of the LBM implementation include:
   • Variable Relaxation Rate: The relaxation rate is updated dynamically based on the local shear-strain rate to accommodate the non-Newtonian viscosity.
   • Porous Medium Treatment: The method handles the quadratic dependence of the drag force on velocity. For isotropic permeability, an explicit velocity solution is derived. For anisotropic permeability, a fixed-point iteration scheme is proposed, though the authors note this significantly increases computational cost.
   • Forcing Scheme: The Guo forcing scheme is utilized to incorporate body forces, and a specific collision term accounts for spatially varying porosity.
4. Simulation Setup: The study utilizes an anonymized CT scan of a patient's aneurysm. Coiling procedures are simulated using a discrete elastic rod model to generate coil configurations with packing densities of 15%, 20%, and 25%. The resulting coil distributions are converted into porosity fields via windowed convolution. Boundary conditions include no-slip walls, a Hagen–Poiseuille inflow profile derived from a 1D-0D network model, and an extrapolation-based outflow.

Key Contributions

• Integrated Workflow: The paper presents a unified framework that combines LBM, VANSE, and non-Newtonian rheology to simulate coiled aneurysms without requiring the explicit discretization of every coil wire.
• Validation of Volume-Averaging: The study provides a direct comparison between fully resolved coil geometries and the volume-averaged porous medium approach.
• Patient-Specific Application: The method is applied to a realistic, patient-specific geometry derived from clinical CT data, demonstrating the feasibility of assessing treatment efficacy on a case-by-case basis.

Results

• Flow Reduction: Simulations confirm that coil insertion significantly reduces flow velocity and momentum within the aneurysm sac while maintaining flow in the parent vessel. The volume-averaged approach yields results in good agreement with fully resolved simulations, capturing the overall flow behavior despite losing local fine-scale effects.
• Packing Density Impact: Increasing the packing density (from 15% to 25%) leads to a more pronounced reduction in flow magnitude.
• Wall Shear Stress (WSS): The treatment drastically reduces the wall shear stress magnitude. Untreated aneurysms exhibited WSS magnitudes around 2 Pa, whereas coiled aneurysms showed magnitudes below 1.5 Pa, suggesting a reduced risk of rupture. The difference in WSS between the different packing densities was found to be slight, potentially due to variations in coil placement.
• Quantitative Metrics: Over the simulated cardiac cycle, the volume-averaged model showed a flow magnitude reduction of approximately 50% and a WSS reduction of about 40% compared to the untreated state.
• Computational Efficiency: The volume-averaged approach offers a viable alternative to fully resolved simulations, which are computationally intensive (approx. 15.5 hours vs. 11.1 hours for the specific setup on the tested hardware, though the fully resolved case had a similar cell count, the complexity of the geometry in the resolved case is inherently higher).

Significance and Claims
The authors claim that this workflow enables a patient-specific assessment of blood flow changes due to potential coil treatments. They assert that the volume-averaged model is valid across the range of packing densities typically used in clinical practice (20–25%).

The paper concludes modestly regarding future directions. It suggests that while accounting for anisotropic permeability could theoretically increase accuracy, the current isotropic assumption is likely sufficient for clinical decision-making given the uncertainties in exact coil placement and the need for rapid treatment assessment. The authors identify detailed parameter calibration, uncertainty quantification, and the analysis of anisotropic permeability extensions as open questions for future research.