Research On Respiratory Motion Modeling And Needle Deflection Compensation In Robot-assisted Liver Puncture Surgery

Percutaneous puncture is an important operation in the diagnosis and treatment of liver cancer.Accurate puncture is the premise to ensure the early diagnosis rate and ablation coverage rate of liver cancer.At present,clinical puncture surgery is performed by doctors under the guidance of computed tomography(CT),magnetic resonance imaging(MRI),ultrasound(US)and other medical images.Due to the movement and deformation of the liver caused by physiological respiration,and intraoperative needle deflection caused by needle-tissue interaction,the accuracy and stability of manual puncture are difficult to be guaranteed.Doctors often have to go through several exploratory repeated operations to accurately insert the needle into the target tumor,and surgical results are highly dependent on doctors’ experience.Robotic puncture can improve the accuracy and stability of puncture surgery while reducing the burden on doctors.However,the existing puncture robots neither solve high-quality,all-round accurate positioning of the intraoperative tumor and surrounding anatomical structures under physiological respiration,nor does they compensate for the needle deflection during the puncture process,therefore further research is needed.In this thesis,aiming at percutaneous liver tumor puncture,in view of the difficulties of clinical puncture surgery and the shortcomings of the existing puncture robots,a preoperative CT-intraoperative ultrasound fusion-guided dual-arm robotic puncture method is proposed.For the robotic puncture scheme,the corresponding puncture procedure is formulated,a corresponding dual-arm robotic puncture system is built,and a software system consisting of an image archiving and communication module,a system calibration module,a navigation module and a robot puncture control module is developed.In particular,in view of the low efficiency and insufficient robustness of the existing ultrasonic probe calibration methods,a calibration method based on the improved N-line model is proposed,and the closed solution of the optimal homogeneous transformation parameters is obtained,finally,the coordinate mapping between image space and robot space is realized.Aiming at the difficulty in locating tumors and surrounding anatomical structures under the influence of physiological respiration,a hierarchical modeling method of personalized respiratory motion is proposed.In order to obtain accurate and personalized prior knowledge of the respiratory liver motion,and to ensure that the internal topology of the liver remains unchanged,a hybrid deformation registration method is proposed,which combines global affine registration and diffeomorphic deformation registration.Aiming at the differences of physiological breathing motions between cycles and within cycles,based on the support vector regression method,a respiratory tumor motion model is established with the motion of body surface markers as input.Based on the acquisition of prior knowledge of the liver motion and the estimation of the tumor motion,with the tumor motion as input,a modeling method of the respiratory liver motion based on the interpolation in Lie algebra space is proposed,which realizes the optimal unbiased estimation of the diffeomorphic dense deformation fields of the liver without the need of intraoperative iterative optimization calculation.Aiming at the challenge of cross-modal and cross-dimensional nature of the CTUS registration,a registration method based on decoupling strategy is proposed.By performing three-dimensional ultrasound(3DUS)imaging on the target,the CTultrasound image registration problem is decomposed into the cross-modal CT-3DUS registration and the cross-dimensional 3DUS-US registration.To circumvent the need for a cross-modal similarity criterion,CT-3DUS registration is performed based on the surface point set of liver vessel.In order to extract the surface point set of hepatic vessels,a method of vessel segmentation in ultrasonic images based on image style transfer and cross-modal transfer learning is proposed,and the problems of insufficient ultrasound image data and difficult labeling are solved.By modeling3DUS-US registration as a maximum posteriori probability state inference problem on a six-node pairwise Markov random field(MRF),a fast and robust global optimal transformation parameter search is realized based on a discrete optimization method.Aiming at the problem of deflection relative to the preoperative planned path caused by the unbalanced force on both sides of the needle tip during the puncture of the bevel-tip needle,a needle deflection compensation method is studied.Firstly,on the basis of the force analysis of the needle-tissue interaction process,by considering the needle as a cantilever beam under the action of the lateral force of the needle tip and the supporting force of the surrounding soft tissue,a needle flexural deformation model is established,and the preoperative prediction of needle flexural deformation is achieved under the premise of known geometry and material properties of the tissue and needle.Then,based on the analysis of the mechanism of the insertion angle shift at the tissue interface,the insertion angle shift experiment is designed and carried out,and the empirical formula of the insertion angle shift is obtained.On the basis of intraoperative prediction of needle deflection,a multi-criteria optimal puncture path planning method based on the weighted sum of cost functions is proposed.The planned puncture path not only compensates for the needle flexural deformation and the insertion angle shift,but also realizes the optimal compromise of various evaluation indicators including the distance from the key anatomical structures,the needle insertion length,and the insertion angle,which improves the accuracy and safety of puncture surgery.For the method proposed in this thesis and the built preoperative CTintraoperative ultrasound fusion-guided dual-arm robotic puncture system,the corresponding phantom experiments and animal experiments are designed and carried out,and the effectiveness and safety of the proposed method and the built system are verified.The research of this topic will promote the development of puncture surgery robots and related technologies,and provide theoretical and technical basics for the realization of safe and reliable high-precision diagnosis and treatment of liver cancer.