| Robot-assisted system for minimally invasive spine surgery is emerging in recent years to help doctors to efficiently complete the operation. Due to the capability to accurately locate the target point and the high precision for positioning the robotic arm, it is used more and more widely for puncture operation. In general, a typical robot-assisted system is composed of three parts:(1) computed tomography (CT) devices; (2) computer assisted data processing and navigation system; (3) robotic arm and con-trol system. Specifically, the second part is further divided into the following modules: image capturing and processing system, virtual surgery simulation system, navigation and control system, and graphical user interface. The work of this dissertation focuses on the "image processing" and "virtual surgery simulation system".Both medical image processing and virtual surgery simulation have important sig-nificance for the robot system. With the aid of image processing techniques, metal artifacts reduction and medical image segmentation can be conducted. Consequently, the quality of medical images can be improved to assist surgeons to find out the target points more accurately. In addition, virtual surgery simulation system builds a virtual surgery environment, in which virtual surgery can be performed to help surgeons judge whether the operation is safe. According to these two modules, the main contents of this dissertation are as follows:(1) Research on metal artifacts reduction algorithms in CT images. In minimally invasive spine surgery, there often exist metal artifacts around the surgery tool due to the high attenuation of X-ray. Based on the principle of computed tomography, the concept of re-rectification is proposed to reduce these artifacts. Specifically, through the processing of "projection-interpolation-back projection", shadow around the tool is attenuated efficiently. For the interpolation processing, two algorithms are presented in this thesis-cubic polynomial interpolation algorithm and nonlinear weight regulation algorithm. Subsequently, experiments are performed on synthetic images with the two new interpolation algorithms to verify their efficiency. Furthermore, these two inter-polation methods are employed in the re-rectification algorithm, and experiments are performed on real medical images. The experiment results show good performance for real images as well as the synthetic images. Considering the fact that the metal surgery tool exhibits different widths from different projection angles, the experiment results of the latter interpolation approach are obviously better than that of "cubic polynomial interpolation algorithm".(2) Research on 2D/3D medical image segmentation algorithms based on multi-threshold single level set method. A multi-threshold single level set model is proposed based on Chan-Vese model. After iterative updating for the single level set function, a medical image can be segmented to multiple regions by different thresholds exactly and efficiently according to intensities. It should be noted that bones which are the most interested region can be segmented clearly. The results of applying this model on different synthetic and real medical images approve the efficiency of the model. By choosing parameters properly, the algorithm performs 2D/3D medical image segmen-tation efficiently. The presented algorithm gets over the disadvantage that only two regions can be distinguished by the Chan-Vese model. Thus, with this model, it is able to finish 2D/3D medical image segmentation quickly and pick up the contours of bones reasonably. As a result of the new detection technique, surgeons can find out target point of the bones more quickly. As an important reference to diagnose and surgical planning, it is meaningful for the surgery.(3) Research on collision detection algorithms between surgical tool and spine 3D model. Collision detection is a critical problem in virtual surgery simulation system. In this research, a collision detection algorithm based on distance field is developed to provide an effective solution. Firstly, EVDT (Effective Vector Distance Transform) algorithm and common distance computation algorithm are combined to compute the signed distance field matrix of spine 3D model. Common distance computation al-gorithm can calculate the distance from a point to a triangle exactly. So computation precision is improved by combining with this algorithm. With this signed distance field, the shortest distance between surgical tool and spine 3D model can be calculated, which can be utilized with sign together to judge whether a collision happens. Furthermore, the safety of surgical solutions can be evaluated according the detection result.(4) Design and implementation of virtual surgery simulation system. Spine 3D model is built with Marching Cubes algorithm from medical image series. Besides, the 3D model of surgical tool is added to the virtual environment. Virtual surgery simulation system is developed with Visual Studio 2005 and OSG(Open Scene Graph) toolkit. With its help, it is able to show the surgery solution to surgeons directly and provide graphical user interface. With the signed distance field matrix which has been computed already, collision detection between surgery tool and spine 3D model can be performed. Hence, dangers can be detected in advanced. As a second pair of eyes of surgeons, it is effective to help judge the feasibility and the safety of a surgical solution.Finally, the research results of this dissertation are combined with mechanical and control system to construct a robot-assisted system for minimally invasive spine surgery. With this robotic system, our research group has performed several experi-ments of minimally invasive spine surgery on man-made spine model and living ani-mal. The results demonstrate that the virtual surgery simulation system facilitate the surgeries effectively and enhance the operation accuracy remarkably.
|
PDF Full Text Request