Mechanical and Civil Engineering Seminar

Title: "Autonomous Robotic Endovascular Navigation via Imitation Learning"

Abstract:

In endovascular surgery, interventionists push a catheter, guided by a thin wire, to reach a treatment site inside the patient's blood vessels to treat various conditions such as blood clots, aneurysms, and malformations. With standard non-robotic tools, interventionists often encounter hazardous navigation challenges. Robotic catheters enhance maneuverability, but they increase the complexity of tool control and the clinical workflow. Autonomous robotic catheter navigation has the potential to overcome these challenges and increase the precision and safety of these procedures. To this end, we have implemented a goal-conditioned imitation learning framework adapted to the unique challenges of endovascular surgery, which include controlling highly underactuated tools through partial observability and variable anatomy. Preliminary results on the benchtop demonstrate the promise of our technique, outperforming state-of-the-art learning policies and classical controllers, while generalizing to unseen geometries. These insights provide an important next step toward leveraging the benefits of end-to-end imitation learning to the benefit of patients and clinicians.

Bio:

I am a PhD candidate in Mechanical Engineering and NSF Graduate Research Fellow at Johns Hopkins University, advised by Professors Axel Krieger and Jeremy Brown. My research lies at the intersection of novel tool design, control, and AI-enabled autonomy for minimally invasive surgery. My focus is developing soft and continuum robotic catheters as well as autonomous navigation frameworks for the endovascular treatment of aneurysms, blood clots, and other vascular diseases. I obtained a MSE in Robotics from Johns Hopkins University in 2025, and a BS in Mechanical Engineering with minor in Control and Dynamical Systems from Caltech in 2022.