Search for: All records

Safely and accurately navigating needles percutaneously or endoscopically to sites deep within the body is essential for many medical procedures, from biopsies to localized drug deliveries to tumor ablations. The advent of image guidance decades ago gave physicians information about the patient’s anatomy. We are now entering the era of AI (artificial intelligence) guidance, where AI can automatically analyze images, identify targets and obstacles, compute safe trajectories, and autonomously navigate a needle to a site with unprecedented accuracy and precision. We survey recent advances in the building blocks of AI guidance for medical needle deployment robots (perceiving anatomy, planning motions, perceiving instrument state, and performing motions) and discuss research opportunities to maximize the benefits of AI guidance for patient care. 

Tendon-driven continuum robot kinematic models are frequently computationally expensive, inaccurate due to unmodeled effects, or both. In particular, unmodeled effects produce uncertainties that arise during the robot’s operation that lead to variability in the resulting geometry. We propose a novel solution to these issues through the development of a Gaussian mixture kinematic model. We train a mixture density network to output a Gaussian mixture model representation of the robot geometry given the current tendon displacements. This model computes a probability distribution that is more representative of the true distribution of geometries at a given configuration than a model that outputs a single geometry, while also reducing the computation time. We demonstrate uses of this model through both a trajectory optimization method that explicitly reasons about the workspace uncertainty to minimize the probability of collision and an inverse kinematics method that maximizes the likelihood of occupying a desired geometry. 

Autonomous robotic inspection, where a robot moves through its environment and inspects points of interest, has applications in industrial settings, structural health monitoring, and medicine. Planning the paths for a robot to safely and efficiently perform such an inspection is an extremely difficult algorithmic challenge. In this work we consider an abstraction of the inspection planning problem which we term Graph Inspection. We give two exact algorithms for this problem, using dynamic programming and integer linear programming. We analyze the performance of these methods, and present multiple approaches to achieve scalability. We demonstrate significant improvement both in path weight and inspection coverage over a state-of-the-art approach on two robotics tasks in simulation, a bridge inspection task by a UAV and a surgical inspection task using a medical robot.