Ann Majewicz, PhD1, Meenakshi Narayan1, Michael Choti, MD, MBA2. 1University of Texas at Dallas, 2UT Southwestern Medical Center
In robotic needle steering, long, thin, flexible needles with asymmetric tips can be steered to reach challenging targets and curve around obstacles. This can be particularly useful for medical procedures in which needle insertion deep within an organ is required, such as biopsy, radiofrequency ablation, and brachytherapy. Due to the flexible nature of the needles, and the inherent inhomogeneity of the tissue, adverse events may occur. These events include needle buckling, undesired needle curvature, and tissue displacement. These adverse events could lead to detrimental steering outcomes such as erroneous needle placement or damage to patient tissue.
We have developed data-driven algorithms to detect various adverse needle steering events. Our sensing system consists of a steerable needle with an embedded electromagnetic position tracker, a force-torque sensor, mounted on the base of the needle, and an electromagnetic position tracker attached to the surface of the tissue. The algorithms could be implemented in real-time, and only require current and prior information regarding needle and tissue position, and force.
To evaluate the algorithms, we developed artificial tissues designed specifically to elicit needle buckling, tissue pushing, or curvature changing behavior. We then inserted needles into the tissues at different insertion velocities (1, 2.5, and 5 mm/s) and with different curvature. Curvature was controlled by duty-cycle spinning the needle to range between fully curved (0% duty cycle) and fully straight (100%) in 25% increments. Our detection algorithms successfully detected the adverse behavior in all cases. Additionally, we observed an new type of buckling event which consists of needle sliding along an obstacle after initial buckling. This event typically occurs at low duty cycles and higher insertion velocities. We modified the detect algorithm to also account for this behavior.
Finally, we experimentally validated our methods by conducting several needle insertions into biological tissue (ex vivo animal liver) under live fluoroscopic imaging. We were able to observe and detect all adverse events in the biological tissue, and the timing of the detection correlated with movements observed in the fluoroscopic video.
In future work, we will continue to improve our algorithms to include additional functionality, such as the directionality of adverse events, and will develop image-based methods for detecting tissue displacement. We will also integrate these detection algorithms with an existing teleoperated needle steering robot. Determining exactly how to respond to these events, either by communicating events to the human operator, or autonomously triggering avoidance control subroutines, will be an interesting topic for future work – one that requires careful thought regarding appropriate roles and responsibilities between human operators and surgical robotic systems.
This work has also been submitted to the International Journal of Medical Robotics Research but will not be published by the SAGES conference. If accepted, publication will be in June 2016.