Study Of Markerless Point Registration Technology For Craniotomy Navigation System

The application of neurosurgical craniotomy navigation makes the surgical process more minimally invasive and precise.At present,mainstream surgical navigation uses marker-assisted methods to register patients,which is prone to undesirable situations such as occlusion,falling of markers and affecting the operation space.In order to make up for the insufficiency of the registration method of marked-point surgical navigation,this paper studies the technology of non-marked-point surgical navigation registration.Therefore,this article takes the craniotomy as the starting point,and aims to achieve accurate patient registration,build an experimental platform for the registration of unmarked surgery navigation,study the extraction and processing technology of 3D point cloud,multimodal medical image fusion and reconstruction technology,3D Point cloud space alignment and other technologies.The main research contents of this paper are summarized as follows:First,in accordance with the procedures and requirements of the surgical navigation system,an experimental platform for registration of patients with non-marked point craniotomy navigation is established,and detailed research is made on the solution methods of point cloud alignment,rotation and translation transformation matrix and coordinate transformation parameters in the basic registration theory.Second,combine the 3D printed head model to obtain the actual spatial point cloud of the patient,and perform filtering and down-sampling preprocessing in PCL.Based on the characteristics of multimodal medical images and the theory of fusion reconstruction,the patient’s CT and MRI sequences are image fused,and on this basis,high-precision three-dimensional reconstruction of the patient virtual model is completed,and the patient ’ s virtual space point cloud is extracted in PCL And pre-processing.Third,aiming at the problem of poor initial point cloud alignment during surgical navigation registration,the key point detection and multiple feature descriptors were separately fused in the initial alignment algorithm,and Kd Tree was used to join the point-pair search process.Based on the SAC-IA algorithm,three types were proposed.An initial alignment method for point clouds based on different feature descriptions.The methods are evaluated through experiments to determine the most suitable initial alignment method for surgical navigation registration.Finally,aiming at the deviations in the initial alignment and various shortcomings of the traditional ICP algorithm,this paper proposes an improved accurate alignment method.Continue to use Kd Tree to speed up the search for corresponding point pairs in the precise alignment process,and use the normal vector included angle threshold to eliminate mismatched points,replace the original error function with the target symmetric error function to enhance the robustness of the ICP algorithm,and finally Design experiments verify the effectiveness and applicability of the algorithm in this paper.