Predicting post-TEVAR endoleaks: a pre-operative hemodynamic risk factor from patient-specific Fluid-Structure Interaction simulations

This study proposes a new pre-operative hemodynamic risk factor, derived from patient-specific Fluid-Structure Interaction simulations, to effectively predict the occurrence of type I and III endoleaks following Thoracic Endovascular Aortic Repair (TEVAR).

The Gist

Imagine your body's main highway, the aorta, develops a weak, bulging spot called an aneurysm. It's like a tire that's about to blow out. To fix this, doctors perform a procedure called TEVAR. Think of this as slipping a sturdy, expandable tube (a stent-graft) inside the highway to bypass the weak spot and keep blood flowing safely.

However, just like putting a new liner in a pipe, sometimes the fit isn't perfect. Blood can leak around the edges of the tube or through gaps in the tube itself. These leaks are called Endoleaks, and they are dangerous because they keep pressure on the weak spot, risking a rupture.

This paper is about a team of engineers and doctors who wanted to build a "Crystal Ball" to predict these leaks before the surgery even happens.

The Problem: Guessing the Fit

Currently, surgeons look at CT scans (3D pictures of the heart) and use their experience to decide where to place the tube. They try to find a "Landing Zone"—a stretch of healthy, straight pipe where the tube can seal tight. But the human body is messy. The aorta twists, turns, and curves. Sometimes, even if the surgeon picks what looks like a good spot, the blood flow is so turbulent that it pushes the tube off the wall, causing a leak.

The Solution: A Digital Twin

The researchers created a digital twin of ten different patients' aortas. They didn't just look at the shape; they ran a super-complex video game simulation called Fluid-Structure Interaction (FSI).

• The Analogy: Imagine blowing air through a flexible, wiggly straw. If you blow too hard or at the wrong angle, the straw might collapse or shift.
• The Simulation: Their computer model simulated blood (the fluid) pushing against the aortic wall (the structure) with incredible detail. They calculated exactly how much "drag" or pushing force the blood exerts on the walls where the tube would sit.

The New "Risk Factor" (The Crystal Ball)

From these simulations, they invented a new score called Risk Factor R. Think of this score like a weather forecast for your surgery:

• Low Risk (Green Light): The blood flow is calm and pushes evenly against the wall. The tube will likely stay put.
• Moderate Risk (Yellow Light): The flow is a bit tricky. It might hold, but it's risky.
• High Risk (Red Light): The blood is slamming into the wall at a bad angle or with too much force. The tube is likely to be pushed away, causing a leak.

They tested this score on 10 patients. They split them into two groups:

1. The Calibration Group (8 patients): They used these cases to teach the computer what a "bad score" looks like. They found that if the score was above a certain number, a leak usually happened.
2. The Validation Group (2 patients): They tested the computer's prediction on these two patients without knowing the outcome first.

The Results: It Worked!

The "Crystal Ball" was surprisingly accurate:

• For the patient who developed a leak, the computer had flagged the specific spot as High Risk before the surgery even started.
• For the patient who had no leaks, the computer correctly predicted Low Risk everywhere.

Why This Matters

This is like having a GPS that doesn't just show you the road, but tells you exactly where the potholes are before you drive over them.

• For Surgeons: Instead of guessing where to place the tube, they can run this simulation first. If the "Risk Score" is high in the spot they planned to use, they can move the landing zone to a safer, calmer spot.
• For Patients: This could mean fewer complications, fewer follow-up surgeries, and a much safer procedure.

The Catch

The researchers admit this is just the beginning. They only tested 10 people, and their computer models used some standard assumptions (like using a generic blood pressure wave instead of a patient's specific one). But, it's a powerful first step toward using math and physics to make life-saving surgeries safer and more predictable.

In short: They turned complex fluid dynamics into a simple "Risk Score" that helps doctors choose the perfect spot to seal a leaky aorta, potentially saving lives by preventing leaks before they happen.

Technical Summary

1. Problem Statement

Thoracic Endovascular Aortic Repair (TEVAR) is a standard minimally invasive treatment for Thoracic Aortic Aneurysms (TAA). However, a significant limitation is the risk of Endoleaks (ELs), specifically Type I (inadequate seal at the proximal or distal landing zones) and Type III (structural failure or separation of modular stent-graft components). These complications often necessitate re-intervention.

Current pre-operative planning relies heavily on anatomical metrics (e.g., landing zone length, aortic tortuosity) but lacks a quantitative, patient-specific assessment of the hemodynamic forces acting on the aortic wall prior to stent deployment. While Computational Fluid Dynamics (CFD) and structural analysis have been used separately, there is a gap in utilizing Fluid-Structure Interaction (FSI) to quantify drag forces and predict post-TEVAR complications.

2. Methodology

A. Patient Cohort and Data

• Subjects: 10 patient-specific TAA cases treated with TEVAR.
• Data Source: Pre-operative, post-operative, and follow-up angiographic CT scans (mean follow-up: 1.5 years).
• Outcomes:
  • Calibration Set (8 patients): 4 with no complications, 4 with ELs (Type IA, IB, or III).
  • Validation Set (2 patients): 1 with no complications, 1 with Type IB EL.
• Anatomy: Landing zones (LZs) were defined as Proximal (PLZ), Distal (DLZ), and Overlap (OLZ) zones.

B. Computational Framework (FSI)

The study employed a high-fidelity Fluid-Structure Interaction (FSI) simulation framework:

• Geometry Reconstruction: Patient-specific aortic lumens were segmented from pre-TEVAR CT scans (from Ascending Thoracic Aorta to the celiac trunk).
• Mesh Generation:
  • Fluid Mesh: Tetrahedral elements with a boundary layer (3 sub-layers).
  • Structural Mesh: Extruded from the fluid surface, modeled as a three-layered solid. Wall thickness was 10% of the radius for healthy tissue and fixed at 1.5 mm for aneurysmal regions.
• Physics Models:
  • Fluid: Newtonian, incompressible blood modeled via Navier-Stokes equations in Arbitrary Lagrangian-Eulerian (ALE) formulation. Turbulence was modeled using the $\sigma$-model Large Eddy Simulation (LES).
  • Structure: Linear elasticity. Young's modulus set to 0.8 MPa (healthy) and 1.2 MPa (aneurysmal).
• Boundary Conditions (BCs):
  • Inlet: Physiological time-dependent pressure wave (70–120 mmHg).
  • Outlets: 3-element Windkessel model for the descending aorta; resistance models for supra-aortic vessels.
  • Structural: Fixed displacement at inlets/outlets; Robin condition (springs) on the external surface to simulate tissue constraints.
• Solver: Simulations ran for 10 cardiac cycles (first 3 discarded for stabilization) using the lifex C++ library on the LEONARDO supercomputer.

C. Development of the Risk Factor ($R$)

The core innovation is a novel hemodynamic risk factor, $R$, designed to predict EL formation based on pre-operative drag forces.

1. Drag Force Calculation ($DF$): Computed at the systolic peak for specific Landing Zone (LZ) surface areas ($A$):
   $$DF = \frac{1}{A} \left( \int_A p \mathbf{n} \, dx + \int_A \mathbf{WSS} \, dx \right)$$
   Where $p$ is pressure, $\mathbf{n}$ is the outward normal, and $\mathbf{WSS}$ is the wall shear stress vector.
2. Component Analysis:
   • $DF_\perp$ (Perpendicular): Hypothesized to cause stent-graft detachment from the opposite wall (Type I EL) or excessive vessel dilation.
   • $DF_\parallel$ (Parallel): Hypothesized to cause device migration (Type III EL).
3. Tortuosity Probability ($p_{tortuosity}$): A statistical probability function derived from literature (Ueda et al.) linking LZ tortuosity to EL risk.
4. Local Risk Factor ($r$): Calculated for 2-cm initial segments of LZs:
   $$r = \frac{DF_\perp^2 \cdot DF_\parallel \cdot p_{tortuosity}}{10^6}$$
   Note: $DF_\perp$ is squared to emphasize its impact on both Type I and III ELs.
5. Normalization to Global Risk ($R$): The local $r$ values are scaled to the interval $[0, 1]$ using the minimum and maximum $r$ values found across all segments for a specific patient:
   $$R = \frac{r - r_{min}}{r_{max} - r_{min}}$$

3. Key Contributions

• First FSI-based Drag Force Analysis for TEVAR: Unlike previous studies using static CFD or structural analysis alone, this work uses FSI to capture the interaction between blood flow and the deformable aortic wall to quantify drag forces.
• Novel Predictive Metric ($R$): Introduction of a dimensionless risk factor that integrates drag force magnitude, direction (parallel vs. perpendicular), and anatomical tortuosity.
• Segmental Analysis: The method evaluates risk not just for the whole landing zone, but for incremental 0.5-cm segments, allowing surgeons to identify specific "hot spots" of high risk within a chosen landing zone.
• Risk Stratification: Established thresholds for risk levels:
  • Low Risk: $R \le 0.33$
  • Moderate Risk: $0.33 < R \le 0.67$
  • High Risk: $R > 0.67$

4. Results

Calibration Set (8 Patients)

• Correlation: All Landing Zones (LZs) that developed Endoleaks exhibited $R > 0.9$ (High Risk), with one minor exception ($R=0.71$).
• Negative Cases: Most LZs without complications had $R < 0.20$.
• Exceptions: Two patients (PT-09, PT-22) had moderate $R$ values (0.62 and 0.39) in the PLZ despite no ELs. The authors attribute this to the high curvature of the Aortic Arch (Zone 3), suggesting the model correctly identifies "hostile" hemodynamics even if the seal held in these specific cases.
• Trend: Extending the Distal Landing Zone (DLZ) away from the aneurysm generally reduced $R$. However, extending the Proximal Landing Zone (PLZ) did not always reduce risk due to the inherently hostile hemodynamics of the aortic arch.

Validation Set (2 Patients)

• PT-67 (Developed Type IB EL): The model correctly predicted a High Risk ($R=1.00$) for the Distal Landing Zone (DLZ) where the leak occurred. Other zones were predicted as Low Risk, matching the clinical outcome.
• PT-40 (No Complications): The model predicted Low Risk for all zones, correctly identifying the patient as low-risk.
• Overall Accuracy: The risk factor successfully identified all observed Endoleak cases and the specific zones involved.

5. Significance and Limitations

Significance:

• Pre-operative Planning: This tool offers a quantitative method to evaluate the safety of proposed Landing Zones before surgery. It can help clinicians decide whether a specific landing zone is hemodynamically hostile and if extending the zone or changing the stent configuration is necessary.
• Mechanistic Insight: It validates the hypothesis that pre-operative drag forces and tortuosity are strong predictors of post-operative stent-graft failure.
• Clinical Impact: By identifying high-risk segments, surgeons can potentially reduce the rate of re-intervention and improve long-term TEVAR outcomes.

Limitations:

• Sample Size: The study is based on only 10 patients, which is insufficient for statistically robust generalization but adequate for a preliminary exploratory study.
• Boundary Conditions: The inlet pressure wave was generic (from literature) rather than patient-specific due to the retrospective nature of the study.
• Material Modeling: The aortic wall was modeled as linear elastic, whereas real tissue is hyperelastic and viscoelastic. However, the authors argue this is acceptable since the risk factor relies primarily on fluid forces.
• Structural Factors: The model focuses on hemodynamics and does not currently incorporate structural stress metrics (e.g., von Mises stress) or specific stent-graft mechanical properties.

Future Work: The authors plan to expand the patient cohort, conduct prospective studies, and integrate patient-specific inlet conditions and structural stress analysis to refine the predictive model.