Vision Servoing Endovascular Interventional Robot System: Design and Verification

In vascular interventions, visual feedback based on DSA images is essential for making interventional action decisions. Real-time DSA images provide crucial feedback on the position of the guidewire tip within the vasculature. When excessive bending of the guidewire tip is detected, interventional radiologists manipulate the guidewire by retracting or rotating it to prevent damage to the intimal surface. This paper aims to integrate visual feedback into the closed-loop control of vascular robots, creating a safe and intelligent vascular interventional robot system. We propose a distributed hardware and software architecture for the vascular interventional robot, building on our previous designs. This architecture allows the robot to support and control various combinations of instruments, ensuring effective data and command transmission between modules. Based on this foundation, we introduce a vision servoing robot control strategy. The microwire target in the DSA image is segmented, and the bending energy is calculated based on the wire’s curve. This energy is then nonlinearly constrained to form a closed-loop control logic. We deployed the vascular interventional robot system in a standard interventional operating room and conducted experiments on vascular phantoms. These experiments verified the robot’s teleoperation functionality, communication latency between modules, and the enhancement of procedural safety and intelligence through vision servoing. The quantitative evaluation indicated that the vision servoing function significantly mitigates the bending of the microwire tip, enabling a smoother completion of the interventional task.