Soft robotic manipulation for intraoperative MRI-guided non-contact laser surgery

Abstract

Magnetic resonance imaging (MRI) offers high-contrast soft tissue imaging and the unique capability of temperature sensing inside the tissue, making it superior guidance in image-guided surgery. Laser-based tumor ablation is one of the treatments that has significantly benefited from MRI guidance, with which 3-dimensional (3D) ablation margins alongside thermal distributions can be evaluated in real-time to protect surrounding critical structures while ensuring adequate ablation margins. However, the confined bore of MRI and its high magnetic field significantly limit surgeons' access to patients, which facilitates the development of MRI-guided robotic systems to allow remote control. Although many systems were proposed to enable intra-operative MRI-guided robotic laser ablation, they generally only allow the targeting of rigid and contact-based laser probes inserted to the surgical site along a straight pathway. There is still a gap for developing an MRI-guided robotic system capable of flexible navigation to targeted lesions and performing delicate non-contact laser beam manipulation.

The main focus of this thesis is to develop a robotic system that integrates a soft continuum manipulator and its high-precision control to enable flexible and precise laser beam manipulation in MRI. Soft robots, benefiting from their elastic body, ensure safe interaction with their surroundings, thus allowing noninvasive and flexible access to the deep surgical sites through confined natural orifices, e.g., transoral approach. Soft robots can be fabricated using magnetic resonance (MR) safe materials and driven by pressurized fluid flow, which brings new opportunities to the development of MRI-guided robotics. The challenge of precise motion control for soft robots is addressed by employing closed-loop control with learning-based modeling and vision feedback. An online learning visual servo control framework is proposed to approximate the nonlinear robot kinematics without prior knowledge of the robot and camera parameters, and enable precise robot navigation even under unknown external disturbances. To achieve delicate laser beam steering using soft robotic manipulators, eye-to-hand visual servo controllers are investigated based on epipolar geometry modeling and machine-learning modeling. A novel miniature soft laser manipulator that ensures MR safety but maintains dexterous manipulation is designed, with its parameters optimized using finite element analysis (FEA). All in all, an integrated soft robotic system is proposed to achieve MRI-guided transoral laser surgery. The robot enables endoscopic laser delivery and operating in the oral and pharyngeal region with sub-millimeter accuracy (<0.2 mm). A patient-specific dental guard is designed to create an open-jaw cavity for robot anchorage and access of auxiliary instruments such as a fiberscope. Novel wireless MR markers are incorporated to enable positional tracking in the MRI coordinate. Furthermore, preclinical trials were conducted to evaluate the robot performance both with ex-vivo swine tissue and a cadaver model.