Bioinspired needle steering: a mechanics-based model

Abstract

Increasing adoption of robotic assisted surgery has opened up the possibility of using novel surgical instruments with control complexity beyond the dexterity of a human user. Steerable needles are one such technology, developed within the research community to enable greater clinical access in minimally invasive surgical procedures.

Of the numerous designs developed, the Programmable Bevel-tip Needle (PBN) is unique in that it does not rely on rotation about its long axis but instead steers via a bioinspired sliding mechanism: relative axial motion of the comprising segments allows the shape of the needle tip to be configured during insertion. Due to interaction with the surrounding tissue, this results in deformation of the needle tip and subsequent deviation from a straight trajectory. This unique steering mechanism allows the needle to be manufactured from a compliant and flexible material, lending itself to intracorporeal medical applications.

The research contributions in this thesis relate to understanding, modelling and evaluating the steering behaviour of the PBN. Previous needle steering models present within the academic literature are presented and discussed in detail. However they are limited in their ability to capture the behaviour of the PBN's steering mechanism due to the highly nonlinear characteristic of the multi-layer design.

The first research contribution of this thesis is a mechanics-based model of the needle's multi-layer structure which considers the force interaction with surrounding tissue. Given a needle tip configuration, this model is able to predict the resulting steering, described by a curvature vector which contains curvature magnitude and directions. This multi-layer model is validated through Finite Element simulations and in-vitro experiments.

The second research contribution is an experimental evaluation of the PBN steering behaviour for different tip configurations. A number of experimental needle insertion trials are conducted in a brain-like tissue phantom with a stereo tracking rig. These experiments (i) quantify the steering ability of the needle through the maximum achievable curvature (ii) demonstrate, for the first time, controllable needle steering of the PBN in three dimensions (iii) evaluate the model's ability to capture the nonlinear behaviour. The developed model is used to make predictions of needle behaviour for different needle design parameters, with studies on number of segments and second moment of area presented.

A further research contribution is an 'inverse model' for the PBN. Initially, two general measures of needle steerablility are proposed, based on the concept of manipulability from robotic manipulator theory. Based on these, a suitable optimisation problem is formulated which is able to find the required needle tip configuration to achieve a desired curvature vector. A comparison of suitable cost functions is performed and the ability to generate input profiles to satisfy predefined needle trajectories is demonstrated.

The final contribution showcases a practical application of the research contributions, through integration of the forward and inverse models into a neurosurgical drug delivery system, named Enhanced Delivery Ecosystem for Neurosurgery 2020 (EDEN2020). EDEN2020 aims to provide the first robotic ecosystem to perform drug delivery and in-situ diagnostics for brain cancer. The models are integrated into the control strategy which allows human-in-the-loop control via a haptic master device. Preliminary results from a series of user-controlled needle steering trials are presented which demonstrate successful targeting in a brain-like tissue phantom.

The thesis concludes with a discussion of these contributions and suggestions for future work.