Deformation Prediction and Autonomous Path Planning for Robot-Assisted Endovascular Interventions

Endovascular interventions, as emerging medical therapies, utilize blood vessels as conduits to access anatomically challenging regions deep within the body. Within endovascular interventions, one of the prominent challenges involves maneuvering the instrument tip by coordinating insertion, retraction, and torque actions at the proximal end of the instrument. This intricate task is hindered by the presence of a complex mapping between input actions and resulting motion, rendering precise control and accurate targeting of the desired area difficult. Thanks to the introduction of robotic assistance and the steerability of robotic catheters, the complexity of endovascular interventions has been mitigated.

The integration of steerable catheters and navigation guidance has the potential to reduce the level of expertise required for endovascular interventions. By leveraging autonomous navigation, path-related complications, such as perforation, embolization, and dissection, arising from excessive interaction forces between interventional tools and the vessels, can be effectively addressed and potentially reduced. Within the context of robotic catheters navigating through narrow, delicate, and deformable vessels, path planning presents significant challenges, particularly under complex operating conditions, stringent safety constraints, and the inherent limitations on catheter steering capability. Furthermore, the intricate interplay between the steerable catheter and vessel walls, coupled with the deformable nature of the vessels, intensifies the complexity of achieving reliable and real-time path planning, rendering it a hard problem to solve.

This dissertation aims to develop a safe, accurate, and efficient path planner for steerable robotic catheters. Firstly, this dissertation provides a systematic literature analysis of path planning techniques, collating the findings from the most significant research contributions in the field employing the PRISMA method. In the first part of this dissertation, a novel path planning approach named BFS-GA is proposed, which effectively adheres to the robot curvature constraint while keeping the catheter's path as close to the vasculature's centerline as possible. This path planner is capable of swiftly calculating obstacle-free trajectories that conform to the patient's vasculature, while incorporating the inherent limitations of the catheter such as maximum curvature.

A major challenge during autonomous navigation in endovascular interventions is the complexity of operating in a deformable but constrained workspace with an instrument. To address this, two methods are proposed in the second part of this dissertation to provide a realistic and dynamic environment for path planning. Specifically, a realistic, auto-adaptive, and visually plausible simulator is developed. This simulator has the capability to accurately predict the interplay between catheters and vessel walls. Additionally, it accounts for the deformable nature of the vessels induced by the cyclic heartbeat motion. In addition, a novel deformable model-to-image registration framework is designed to reconstruct comprehensive intra-operative vessel structures from medical imaging data, while accurately accounting for deformations.

Given the dynamic vascular environments generated as above, a robust path planner named C-GAIL for steerable catheters is proposed in the third part of this dissertation. This path planner ensures higher precision and robustness by accounting for both the deformable properties of vessels and the catheter's steering capabilities. The in-vitro experiments demonstrate that the path generated by the proposed C-GAIL path planner aligns better with the actual steering capability of robotic catheters. Thereafter, the dissertation presents an in-depth exploration of path planning assistance utilizing various interactive modalities based on augmented reality. Three interactive control modalities for steering robotic catheters are introduced, and their impact on human-in-the-loop robot-assisted cardiac catheterization is investigated. The path guidance is facilitated by the previously discussed C-GAIL path planning method. A user study is conducted, which demonstrates the feasibility of harnessing the capabilities of a gaming joystick for catheter teleoperation and the practicality of utilizing a head-mounted display to receive 3D visual feedback.