Modeling for active needle steering in Percutaneous surgery

• Main Author: Yan, KaiGuo
• Other Authors: Ng Wan Sing
• Format: Theses and Dissertations
• Published: 2008
• Subjects: DRNTU::Engineering::Mechanical engineering::Surgical assistive technology
• Online Access: https://hdl.handle.net/10356/5453
• Institution: Nanyang Technological University

Precise needle insertion is very important for a number of percutaneous interventions. Yet it is very difficult to achieve in practice. Errors caused by the target movement and needle deflection have been observed for a long time. Yet to date, there are few effective physical-based needle steering systems existing for correcting the needle deflection when it occurs. In addition, many procedures are currently not amenable to needles because of obstacles, such as bone or sensitive tissues, which lie between feasible entry points and potential targets. Thus, there is a clear motivation for needle steering in order to provide accurate and dexterous targeting. This work aims to build a needle-tissue interaction model that can be used for designing a needle steering system. A spring-beam-damper model is adopted in this work, which takes into consideration both the elastic, viscoelastic tissue reactions and the needle flexibility, as well as their interaction effects. Unconstrained modal analysis method, which is computationally efficient, is adopted to analyze the system dynamics. Considering the tissue inhomogeneity and computational efficiency, depth-varying mean parameters are proposed in this thesis to calculate the spring and damper effects. Local polynomial approximations in finite depth segments are used to approximate the unknown depth-varying mean parameters. Based on these approaches, an online parameter estimator has been designed using modified least square method with forgetting factor. Extensive experiments have been carried out in various types of phantoms to validate the needle steering model with the online parameter estimator. Results have shown that the model can track the needle tip trajectory with RMSE (Root Mean Square Errors) less than 1mm after convergence even in the presence of large noises, which come from the reduced mode model, poor initial estimation and sensor noises etc. The convergence rate will be greatly improved if the needle gets supported.