Estimation, Modeling and Control of Steerable Needles for Transendoscopic Deployment into the Lung
AUTHORS
ABSTRACT
This dissertation focuses on the use of estimation and modeling methods to control ultra-flexible needles deployed through endoscopes for the purposes of targeting nodules in the periphery of the lung. We develop estimation and modeling methods to describe these continuum devices and deploy them accurately to point-targets and along pre-defined trajectories. Lung cancer needs early diagnosis for a maximized chance at curative treatment. Determining diagnosis requires biopsy of a suspected tumor, and surgeons typically take two approaches: transbronchoscopic or transthoracic. Though the transbronchoscopic approach is less invasive, it is difficult to use this method to reach nodules in the periphery of the lung. With a transthoracic approach, on the other hand, complication rates such as pneumothorax increase, making the rate of successful biopsy of these peripheral nodules low. Without successful biopsy there can be no definitive diagnosis, and a “wait and see what happens” strategy is often adopted if biopsy cannot be achieved. There must be a better way to get to these nodules. This work leverages a semi-autonomous surgical robot that deploys continuum devices through a clinical bronchoscope to target these hard-to-reach nodules with minimal risk of pneumothorax. To achieve superior targeting accuracy compared to traditional manual tools, it utilizes mechanically-compliant structures, robotic actuation, image guidance and motion planning. Throughout this dissertation, we describe novel estimation and modeling techniques used for accurate control and deployment of the system and present experimental validation of the robotic approach in ex vivo and in vivo animal models. Though transbronchoscopic surgery is already less invasive than alternatives, all of the components should be as small as possible to reduce trauma to healthy tissue. However, miniaturization comes with trade-offs such as difficult control and reduced sensing. We present learned methods that estimate the missing sensor information to reconstruct the entire state and use this estimate under closed-loop control to accurately deploy the needle stage of the robot. We present experiments validating the targeting accuracy of the controller using the estimate in tissue phantoms and ex vivo tissues. Prior work from our lab presented a new kind of steerable needle leveraging a hinge placed at the needle tip providing increased deflection capabilities, but did not present a model that described the motion of the needle. Herein, we extend this work to characterize the motion of the needle and its tip. Additionally, we conduct an observability analysis of the model parameters, an ablation study to show it is a minimal parameter set, and calibrate an experimental prototype. We show that the model predicts the motion of the needle better than other existing models, while simultaneously predicting the flexure hinge transient motion, which no other models capture. All of this work is aimed at providing surgeons with a better set of tools to target peripheral nodules that need biopsy for definitive cancer diagnosis. We contribute the first ever experiments validating a semi-autonomous steerable needle robot in the lungs of an in vivo animal, present novel estimation methods for predicting needle tip state, and present a new model that describes flexure tip needle motion. With improved, smarter, and more capable tools, we hope to provide surgeons with the capabilities necessary to access, treat and diagnose nodules in all areas of the lung.
Tags: Dissertations 2022