Development of a Shape Memory Alloy Actuated Robotic Catheter for Endocardial Ablation: Modeling, Design Optimization, and Control

Atrial fibrillation is the most common cardiac arrhythmia, afflicting more than 2 million Americans. Symptoms include shortness of breath, fatigue, chest pain, stroke, and even death. Treatment options consist of pharmacological, surgical, and electrophysiological (ablation catheter-based) approaches. The ideal treatment would combine the effectiveness of surgical methods with the minimally invasive attributes of catheter-based approaches. However, commercially available catheters possess a number of limitations that hinder their effectiveness. This dissertation focuses on the design optimization and control of a robotic ablation catheter, internally actuated using shape memory alloys (SMAs), that overcomes many of the limitations of existing ablation catheters.;The robotic ablation catheter is constructed from serially connected bending segments actuated by internal SMA tendons. Each bending segment contains four SMA actuators that contract upon heating and produce bending moments. The multiple actuators and segments provide greater navigability for the physician. Coupled with the catheter's computer-controlled capabilities, this robotic catheter has the potential to improve success rates and reduce procedure times in the treatment of AF, while simultaneously reducing healthcare costs and radiation exposure to patients and medical staff.;The kinematics and inverse kinematics of the robotic catheter are developed in two coordinate systems: three-dimensional Cartesian coordinates and generalized coordinates (catheter bending and rotation angles). Control algorithms are developed based on the generalized coordinates, while catheter tip measurements are made in Cartesian coordinates, motivating the need for transformations between the two.;The catheter's bending mechanics are described using a circular arc model, while SMA actuation is modeled using free energy techniques. Two specific cases are considered: single-tendon SMA actuation and antagonistic SMA actuation. Both cases are modeled using COMSOL Multiphysics Modeling and Simulation Software and are experimentally validated.;Design optimization of the robotic catheter is accomplished using the COMSOL models and genetic algorithms (GAs). The geometry and material properties of each model are parameterized and used as design variables in the GA. Both single-objective and multiobjective cases are considered. The single-objective problem optimizes the catheter's radius of curvature, a measure of its navigability. The multi-objective problem optimizes radius of curvature and "pushability", a quality related to catheter stiffness.;The computationally efficient hysteretic recurrent neural network (HRNN) is implemented into a sliding mode control algorithm for the position control of SMA actuators. The method is derived for a constant stress SMA actuator and demonstrated experimentally. The control algorithm is extended to variable stress SMA actuators, the situation encountered in the robotic catheter. Simulation results are presented for a single SMA actuator, and the feasibility of the approach for antagonistic actuation is discussed.