Design and control of a soft robotic manipulation system for transoral laser microsurgery under magnetic resonance imaging

Abstract

Nasopharyngeal cancer is relatively prevalent in Southeastern China such as Hong Kong. Effective treatment for head and neck cancers (HNCs) is of paramount importance. Traditional open incision approach causes trauma to the patient and induces long recovery time, whereas non-surgical chemoradiotherapy is toxic and degrades swallowing function. Therefore, alternative locoregional transoral therapy is vital for curing HNCs with the least post-surgical trauma and side effects. This thesis presents the design, optimization, fabrication and experimental validation of a magnetic resonance (MR) conditional reinforced soft robot (RSR) for transoral laser dissection. A world-first 3D-printed manipulator for MRI-guided laser surgery is fabricated by PolyJet™ printing technology (Stratasys Ltd., U.S.). Multi-printing technology allows seamless application of strain wrapping constraints to enhance robustness and controllability of RSR. Its miniature (Ø13 mm × 100 mm), compliance design enables RSR to operate within a confined oral, nasopharyngeal (ONP) cavity. This five degree-of-freedom (DoF) robotic system has kinematic redundancy to reach the same target via various paths. RSR is powered by a bio compatible hydraulic fluid to minimize the hysteresis effect. Novel MR-based wireless tracking technique is adopted. Three miniaturized microcircuits are embedded in the patient-specific dental guard, which are tracked under MRI to provide a three-dimensional localization of the robotic system. MR-compatibility test has been done to demonstrate no observable image artifact during simultaneous robot motion. Pre clinical validation has been conducted for the proposed 3D-printed robotic system to perform laser dissection task under MRI. Path-following tests have been performed in free space with an electromagnetic tracking system to evaluate its maneuverability. Sufficient accuracy and repeatability have been demonstrated for precise laser dissection task with mean errors of 0.1500 ± 0.0835 mm and 0.0446 ± 0.0082 mm respectively.