Research On 3D Navigation Technology Of Spinal Surgery Assisted Robot System

The threedimensional navigation technology of the spinal surgery auxiliary robot is to locate and track the location and path of the operation accurately with the aid of twodimensional / threedimensional medical image information and positioning tracking device,so as to assist the doctor in the spine surgery to achieve precise puncture.Compared with traditional spinal surgery based on experience of doctor and navigation method based on 2D medical image,three dimensional navigation technology can help doctors determine the lesion points,make surgical plans more directly,and has higher positioning accuracy than others.At present,all of the spinal surgery navigation systems applied in major hospitals are commercial products of foreign companies.In order to break the monopoly of foreign products,it is necessary to develop a navigation system that can accurately and quickly locate the focal point of the spine.In view of the above problems,this paper studies the key technologies such as coordinate transformation,3D reconstruction and 2D-3D registration in the robot threedimensional navigation system,and builds the navigation experiment platform,carries out the experiments of locating position and attitude,and validates the key technologies.In this paper,the hardware and software of the robot's threedimensional navigation system are first introduced.On the basis of the hardware of the original twodimensional navigation system,the distribution and arrangement of the original bicalibration target are improved.Experimentally verified,the improved calibration target can effectively improve the accuracy of the calibration of Xray image and 3D space.On the software platform,the 3D reconstruction module,the 2D-3D registration module and the surgical path planning module are added to realize the transformation of the surgical path to the threedimensional space coordinate.Then the whole robot threedimensional navigation process is introduced in detail and the position and posture conversion relations between hardware and software modules are analyzed.In order to enable the doctor to observe the lesion area more intuitively and develop a surgical plan,the Marching Cubes algorithm in VTK is used for surface rendering,and the light projection method is used for volume rendering.The Marching Cube algorithm,based on the surface contour threshold,constructs a series of triangular patches in the cube containing the surface threshold information,and uses the gradient vector instead of the normal vector to construct the surface illumination model.The ray projection method,based on all the voxel information in the body data,classifies the data according to the CT value,and syntheses the image forward according to the color value and the opacity of each material,and the volume rendering model is obtained.The results of 3D reconstruction show that the surface rendering method based on sliced contour cannot display the internal structure of the model,but because the volume data is less,the rendering efficiency is higher,so that it is suitable for the process of surgical navigation.And the volume rendering,based on all voxel information,the low rendering efficiency is due to the large amount of data.However,it can show lesion points inside the model and is suitable for preoperative diagnosis.The original twodimensional navigation system has realized the calibration of the threedimensional space of the Xray map and the calibration target.Therefore,in order to realize the conversion from the threedimensional model space to the patient space.In this paper,the 2D-3D registration method based on the CT volume data and the Xray map is studied.Firstly,the common methods of transformation,optimization,interpolation and measurement in the registration framework are analyzed.According to the demand of the subject,the ray projection interpolation method is adopted.By simulating the attenuation of light sources through Xray photographing process,the 3D volume data can be interpolated into two dimensional images.The mutual information measure can obtain the statistical correlation between two systems.The registration accuracy is high when the two images have high coincidence degree.However,because of the large amount of computation,the multi resolution strategy is adopted to decompose the image into subimages with different resolution of N layer.Starting from the image with low resolution at the top of the resolution,registration begins from top to bottom,from coarse to fine.The registration results show that the above 2D-3D registration method can achieve threedimensional to twodimensional transformation,and the obtained registration parameters will be used in the subsequent path planning to the patient space conversion.Finally,a threedimensional robotic navigation platform is set up.Using the 2D-3D registration parameters obtained above,the calibration parameters of the Xray maps and the calibrated target and the data of the NDI,the conversion from the planning path to the robot world coordinate system was achieved.The posture is transmitted to the KUKA robot for locating position and posture experiments.The experimental results show that the root mean square positioning error of the navigation system is 4.36 mm,the maximum value is 5.44 mm,and the root mean square attitude error is 4.29 degrees and the maximum value is 5.53 degrees,which can meet the accuracy requirements of spinal surgery.Finally,the deviation sources and solutions in the experiments are discussed.The experimental results prove the correctness of the coordinate transformation and 2D-3D registration methods used in this paper.