Hybrid Path Planning and Reinforcement Learning-Based Magnetic Guidewire Navigation in Cardiac Intervention

Abstract

Magnetically driven microrobots have shown significant potential in biomedical applications, including endovascular intervention, targeted therapy, and drug delivery, particularly in the treatment of cardiovascular diseases. Conventional vascular interventional procedures typically rely on physicians to manually manipulate guidewires under medical imaging guidance, which often suffer from strong dependence on operator experience, prolonged procedure time, and increased patient discomfort. To address these limitations, this study proposes a magnetic vascular robotic navigation method based on path planning and reinforcement learning. By controlling a robotic manipulator to regulate the external magnetic field, the proposed method enables motion control of a magnetic guidewire during vascular interventions. In the proposed framework, the robotic manipulator serves as the actuation platform, where the path planning algorithm provides global navigation guidance, while the reinforcement learning strategy performs local motion control. In addition, a simplified two-dimensional virtual vascular environment was constructed for algorithm training, and the proposed method was deployed on a robotic manipulator system for experimental validation in a real vascular phantom. Experimental results demonstrate that the proposed method achieves stable guidewire navigation within vascular environments and outperforms conventional control methods in key metrics such as procedure time, thereby validating its feasibility and potential for endovascular intervention applications.