ExoNav II: Design of a Robotic Tool with Follow-the-Leader Motion Capability for Lateral and Ventral Spinal Cord Stimulation (SCS)

TL;DR

The paper tackles the challenge of steering spinal cord stimulation electrodes toward ventral and lateral motor fibers to support gait restoration after spinal cord injury. It introduces ExoNav II, a tendon-driven continuum robot with a helically patterned nitinol inner tube inside a 3D-printed outer tube, and a back-end actuation unit that translates, rotates, and tensions a tendon to realize follow-the-leader motion. A forward kinematic model links tendon stroke, inner-tube geometry, and end-effector trajectory, and experimental results show that a position-based estimation yields higher accuracy than a stroke-based one, with RMSE around $8$–$9$ mm and clear FTL capability demonstrated in a spinal cord phantom. These findings indicate the potential for precise, repeatable navigation to ventral/lateral spinal regions in a minimally invasive epidural approach, with future work focusing on closed-loop control and electrode integration for clinical translation.

Abstract

Spinal cord stimulation (SCS) electrodes are traditionally placed in the dorsal epidural space to stimulate the dorsal column fibers for pain therapy. Recently, SCS has gained attention in restoring gait. However, the motor fibers triggering locomotion are located in the ventral and lateral spinal cord. Currently, SCS electrodes are steered manually, making it difficult to navigate them to the lateral and ventral motor fibers in the spinal cord. In this work, we propose a helically micro-machined continuum robot that can bend in a helical shape when subjected to actuation tendon forces. Using a stiff outer tube and adding translational and rotational degrees of freedom, this helical continuum robot can perform follow-the-leader (FTL) motion. We propose a kinematic model to relate tendon stroke and geometric parameters of the robot's helical shape to its acquired trajectory and end-effector position. We evaluate the proposed kinematic model and the robot's FTL motion capability experimentally. The stroke-based method, which links tendon stroke values to the robot's shape, showed inaccuracies with a 19.84 mm deviation and an RMSE of 14.42 mm for 63.6 mm of robot's length bending. The position-based method, using kinematic equations to map joint space to task space, performed better with a 10.54 mm deviation and an RMSE of 8.04 mm. Follow-the-leader experiments showed deviations of 11.24 mm and 7.32 mm, with RMSE values of 8.67 mm and 5.18 mm for the stroke-based and position-based methods, respectively. Furthermore, end-effector trajectories in two FTL motion trials are compared to confirm the robot's repeatable behavior. Finally, we demonstrate the robot's operation on a 3D-printed spinal cord phantom model.

Figures (9)

• Figure 1: (a) ExoNav robotic system design with inner and outer tubes, the roller gear and linear tendon pulling mechanisms at the back-end (left-to-right), (b) close-up view of the inner tube with helical cut patterns and a 3D-printed cap, utilized for anchoring the actuation tendon, installed at the tip of the robotic tool.
• Figure 2: (a) Section of the inner tube showing the rectangular cut patterns and the offset $a$ between consecutive cuts, (b) Isometric view of a separated segment of the inner tube with neutral axis, central axis, and the nitinol tendon, (c) Top view of the separated segment of the inner tube.
• Figure 3: (a) Actuated inner tube bends in a helical shape and forms an imaginary cylinder, (b) Coordinate frames defined at the tip of the outer tube and at the center point of the bottom circle of the imaginary cylinder, with respective angles between frames highlighted.
• Figure 4: (a) Inner tube bending in a helical shape with optical markers attached on it, (b) Total joint space and joint space trajectory estimations, (c) Comparison of ground truth $\phi$ and model predictions, (d) Model-predicted marker trajectories compared to their actual measured positions.
• Figure 5: Maximum Euclidean distance ($max(d_E)$) and RMSE values corresponding to markers position estimations by stroke-based and position-based methods.