A simulation environment for robot-assisted endovascular interventions

Matteo Pescio, Chenhao Li, Dennis Kundrat, Maura Casadio, Giulio DagninoInt. J. Comput. Assist. Radiol. Surg 2025 · DOI

TL;DR

This paper introduces a novel simulation environment for robot-assisted endovascular interventions, focusing on accurate force and shape sensing. By integrating advanced physical modeling with a realistic phantom, they've created a tool that can help train clinicians and develop better robotic surgical systems for minimally invasive procedures.

What Problem Does This Paper Address?

Cardiovascular diseases are a leading cause of death globally, and while endovascular procedures offer a less invasive alternative to open surgery, they are complex and require high precision. Robotic assistance and advanced navigation software are emerging to improve dexterity and guidance, but developing and training these systems requires realistic simulation environments. Current simulators often lack accurate modeling of the physical interactions between surgical tools and the delicate vascular anatomy, particularly concerning forces and deformations. This gap hinders the development of more sophisticated robotic systems and effective clinical training.

Methods

The researchers developed a model-based simulation environment within the SOFA (Simulation Open Framework Architecture) framework. Key aspects include:

  • Peripheral Catheter Modeling: The catheter was modeled using beam theory, representing it as a series of beam elements with six degrees of freedom. This allows for the simulation of both rigid body transformations and small deformations. The dynamics are governed by Newton's second law of motion.
  • Vascular Phantom Modeling: A silicone phantom of an abdominal aorta with an aneurysm was created. Its mechanical properties were characterized through tensile tests. Various hyperelastic models (neo-Hookean, Mooney-Rivlin, Ogden) were evaluated, with the elastic model ultimately chosen for its balance of accuracy and computational performance in the simulation context.
  • Contact and Interaction Modeling: Realistic interactions between the catheter and the vascular model were established using the Finite Element Method (FEM). A friction-inclusive contact model based on Signorini's contact law and Coulomb's friction law was implemented to account for the dry and textured nature of the phantom. Contact detection was managed using the LocalMinDistance method.
  • Model Validation: The simulation was validated through two experimental tests:
  • Catheter Insertion: Measuring contact forces and positional changes during catheter insertion at controlled speeds.
  • Aortic Deformation: Quantifying static forces exerted by the catheter on the vessel wall and measuring the resulting deformation.

Datasets

Not discussed in the paper. The study utilized a custom-built silicone phantom and experimental data collected from physical tests performed on this phantom.

Key Insights and Ideas

  • Importance of Friction: Neglecting friction in surgical simulations can lead to inaccurate force predictions, especially in the context of catheter-vessel wall interactions.
  • Elastic Model Sufficiency: While more complex hyperelastic models like Ogden and Mooney-Rivlin might better fit stress-strain data in isolation, the simpler elastic model provided a better balance of accuracy and computational efficiency for simulating catheter interactions within the vascular phantom. This is crucial for real-time simulation.
  • Validation is Key: The study emphasizes the critical need for experimental validation of simulation models against real-world physical interactions to ensure their clinical relevance and utility.
  • Potential for Advanced Training: The developed simulation environment, with its focus on force and shape sensing, has the potential to significantly enhance training for minimally invasive surgical procedures, allowing clinicians to develop a better feel for instrument-phantom interactions.

Key Results

  • Catheter Insertion:
    • Experimental tests showed that insertion speed minimally influenced force magnitude but affected contact time.
  • Simulated forces were generally higher than experimentally measured forces.
  • Absolute errors in force prediction were: 0.0058 N (24.29%) in the x-direction, 0.034 N (19.35%) in the y-direction, and 0.0371 N (30.45%) for the total force.
  • Positional accuracy was good, with an average Euclidean tip distance of 3.1 mm and a mean Fréchet distance of 3.6 mm for the proximal segment.
  • Aortic Deformation:
    • The elastic FEM model showed the best performance in simulating anatomical deformation, with errors of 34%, 19%, and 59% across three different force scenarios.
  • The simulation accurately replicated displacement measurements, with simulated displacements for low, intermediate, and high forces being 2.20 mm, 5.24 mm, and 7.80 mm, respectively, compared to experimental values of 3.34 mm, 4.61 mm, and 4.92 mm.

Limitations and Future Work

  • Phantom Fidelity: The mechanical fidelity of the silicone phantom to real human vessels is a limitation. The study used isotropic hyperelastic models, but future work will incorporate anisotropic models for more realistic simulation of blood vessel behavior.
  • Manual Alignment: The initial alignment of the catheter's starting position in the simulation required a manual step, introducing potential for minor inaccuracies.
  • Future Directions:
    • Inclusion of anisotropic hyperelastic models.
  • Integration of imaging-derived patient-specific data.
  • Modeling of dynamic blood flow.
  • Simulation of deformation induced by cardio-respiratory activity.
  • These advancements aim to further improve the simulation's realism and clinical applicability for advanced training environments and robotic system development.

Summary updated

BibTeX
@article{pescio2025simulation,
  author = {Matteo Pescio and Chenhao Li and Dennis Kundrat and Maura Casadio and Giulio Dagnino},
  title = {{A simulation environment for robot-assisted endovascular interventions}},
  journal = {Int. J. Comput. Assist. Radiol. Surg.},
  year = {2025},
  doi = {10.1007/s11548-025-03458-2},
}
Figure