Vision-Based Reasoning with Topology-Encoded Graphs for Anatomical Path Disambiguation in Robot-Assisted Endovascular Navigation
Jiyuan Zhao, Zhengyu Shi, Wentong Tian, Tianliang Yao, Dong Liu, Tao Liu, Yizhe Wu, Peng Qi,,∗
AI summary
Problem
Robotic endovascular navigation relies exclusively on 2D angiography, which lacks spatial context and causes projection-induced ambiguities at vessel bifurcations that misguide automated path planning.
Approach
A two-stage pipeline first segments coronary vessels with a spatial-coordinate attention U-Net, then uses a Graph Attention Network to reason over a topology-encoded vessel graph, distinguishing true anatomical bifurcations from 2D projection artifacts to plan safe guidewire trajectories.
Key results
- SCAR-UNet achieves 93.1% Dice coefficient for vessel segmentation
- GAT-based disambiguation attains 95.0% success and 90.0% target-arrival rates
- Outperforms shortest-path and heuristic planners by 20-35% in success metrics
- Validated on a robotic platform confirming real-time feasibility and robustness
Why it matters
It enables safer and more reliable robotic coronary interventions by overcoming 2D imaging limitations, accelerating the clinical adoption of automated endovascular navigation.
Abstract
Robotic-assisted percutaneous coronary interven- tion (PCI) is constrained by the inherent limitations of 2D Digital Subtraction Angiography (DSA). Unlike physicians, who can directly manipulate guidewires and integrate tactile feedback with their prior anatomical knowledge, teleoperated robotic systems must rely solely on 2D projections. This mode of operation, simultaneously lacking spatial context and tactile sensation, may give rise to projection-induced ambiguities at vascular bifurcations. To address this challenge, we propose a two-stage framework (SCAR-UNet-GAT) for real- time robotic path planning. In the first stage, SCAR-UNet, a spatial-coordinate-attention-regularized U-Net, is employed for accurate coronary vessel segmentation. The integration of multi-level attention mechanisms enhances the delineation of thin, tortuous vessels and improves robustness against imaging noise. From the resulting binary masks, vessel centerlines and bifurcation points are extracted, and geometric descriptors (e.g., branch diameter, intersection angles) are fused with local DSA patches to construct node features. In the second stage, a Graph Attention Network (GAT) reasons over the vessel graph to identify anatomically consistent and clinically feasible trajectories, effectively distinguishing true bifurcations from projection-induced false crossings. On a clinical DSA dataset, SCAR-UNet achieved a Dice coefficient of 93.1%. For path disambiguation, the proposed GAT-based method attained a success rate of 95.0% and a target-arrival success rate of 90.0%, substantially outperforming conventional shortest-path planning (60.0% and 55.0%) and heuristic-based planning (75.0% and 70.0%). Validation on a robotic platform further confirmed the practical feasibility and robustness of the pro- posed framework.