Image-based Intraluminal Contact Force Monitoring in Robotic Vascular Navigation

Masoud Razban, Javad Dargahi, Benoit BouletarXiv 2020 · arXiv

Introduction

Endovascular interventions, such as carotid artery stenting, are critical procedures that involve navigating flexible guidewires through complex vascular networks. However, these procedures carry inherent risks, including vessel wall injury, plaque rupture leading to stroke, or vessel perforation. A significant challenge for surgeons is the lack of real-time feedback regarding local contact forces. While clinicians can feel resistance at the tool handle, this "insertion force" does not accurately reflect the localized pressure exerted by the wire's tip or body against the vessel wall. Integrating physical sensors into these micro-scale tools is technically prohibitive and costly. To address this, researchers have developed a "sensor-less" image-based method to monitor these delicate interactions.

Methodology

The proposed system utilizes a three-pillar approach to estimate contact forces without specialized hardware:

1. Real-Time Image Segmentation

Using computer vision techniques, the system extracts vessel boundaries and the guidewire's shape from video frames, mimicking X-ray fluoroscopy. A moving search window algorithm identifies precise contact points where the wire interacts with the vessel wall.

2. Nonlinear Finite Element Modeling (FEM)

The guidewire is represented mathematically using Timoshenko beam theory. This model accounts for the large deflections and shear deformations characteristic of highly flexible surgical instruments. To ensure high fidelity, the researchers mapped the specific flexural rigidity profile of the tools, accounting for varying stiffness from the base to the tip.

3. Inverse Force Estimation

The core innovation lies in using observed visual data as a mechanical input. By measuring the deflection of the guidewire from its neutral state in the images, the system calculates the forces required to produce that specific shape. This allows for the "back-calculation" of reaction forces at every interaction point along the wire.

Cantilever model used to back-calculate contact forces from observed catheter deflection.

Experimental Validation

The system was tested using a transparent, anthropomorphic silicone phantom of the human aortic arch and carotid arteries. A 2-DOF robotic driver navigated a Zipwire™ Stiff guidewire through the phantom, while a 6-DOF force/torque sensor provided ground-truth measurements for comparison.

Key Results and Insights

The study demonstrated that high-risk internal forces often occur even when the insertion force felt by the operator remains low, confirming the necessity of localized monitoring. The highest contact forces (up to 0.26 N) were consistently observed at arterial bifurcations, identifying these as critical "danger zones" for robotic navigation.

Real-time segmentation and contact force mapping on a vascular phantom.

Quantitatively, the system achieved:

  • Estimation Accuracy: Over 87% correlation with physical force sensors.
  • Shape Fidelity: Low FEM estimation error, with a Root Mean Square Error (RMSE) between 1.3 mm and 3.1 mm across different arterial paths.
  • Real-time Mapping: The generation of force "contour maps" that color-code vessel walls based on applied pressure.

Conclusion and Future Work

This "sensor-less" approach demonstrates that high-fidelity safety data can be extracted from standard surgical tools using a combination of computer vision and physics-based modeling. While currently limited to 2D projections and static vessel models, future work will focus on extending the system to 3D environments and incorporating compensation for physiological motion, such as heartbeats and respiration. This technology paves the way for safer, more autonomous robotic vascular interventions.

Summary updated

BibTeX
@misc{razban2020image,
  author = {Masoud Razban and Javad Dargahi and Benoit Boulet},
  title = {{Image-based Intraluminal Contact Force Monitoring in Robotic Vascular Navigation}},
  year = {2020},
  eprint = {2012.10762},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2012.10762},
}
Figure