Full 3D motion control for programmable bevel-tip steerable needles

Abstract

Minimally invasive surgery has been in the focus of many researchers due to its reduced intra- and post-operative risks when compared to an equivalent open surgery approach. In the context of minimally invasive surgery, percutaneous intervention, and particularly, needle insertions, are of great importance in tumour-related therapy and diagnosis. However, needle and tissue deformation occurring during needle insertion often results in misplacement of the needles, which leads to complications, such as unsuccessful treatment and misdiagnosis. To this end, steerable needles have been proposed to compensate for placement errors by allowing curvilinear navigation. A particular type of steerable needle is the programmable bevel-tip steerable needle (PBN), which is a bio-inspired needle that consists of relatively soft and slender segments. Due to its flexibility and bevel-tip segments, it can navigate through 3D curvilinear paths.

In PBNs, steering in a desired direction is performed by actuating particular PBN segments. Therefore, the pose of each segment is needed to ensure that the correct segment is actuated. To this end, in this thesis, proprioceptive sensing methods for PBNs were investigated. Two novel methods, an electromagnetic (EM)-based tip pose estimation method and a fibre Bragg grating (FBG)-based full shape sensing method, were presented and evaluated. The error in position was observed to be less than 1.08 mm and 5.76 mm, with the proposed EM-based tip tracking and FBG-based shape reconstruction methods, respectively.

Moreover, autonomous path-following controllers for PBNs were also investigated. A closed-loop, 3D path-following controller, which was closed via feedback from FBG-inscribed multi-core fibres embedded within the needle, was presented. The nonlinear guidance law, which is a well-known approach for path-following control of aerial vehicles, and active disturbance rejection control (ADRC), which is known for its robustness within hard-to-model environments, were chosen as the control methods. Both linear and nonlinear ADRC were investigated, and the approaches were validated in both ex vivo brain and phantom tissue, with some of the experiments involving moving targets. The tracking error in position was observed to be less than 6.56 mm.