基于增强现实的手术导航图像融合方法研究

发布时间：2018-06-19 04:20

本文选题：增强现实 + 手术导航　；参考：《吉林大学》2015年硕士论文

【摘要】：应用在手术导航系统中的基于增强现实的图像融合技术改变了传统的手术方式，不仅能减轻病人痛苦，有效的降低手术风险，还能精确的制定术前规划，在医疗培训方面也具有很高的应用价值。基于增强现实的图像融合是将虚拟的病灶部位叠加在真实的手术部位，为医生诊断提供更多的支撑。本文研究了二维图像的融合及三维图像的融合可视化，通过对将要手术部位的外表面二维融合和内部生理信息的三维融合，使病灶信息量增大，具体的研究内容和成果如下： 1.本文结合真实的手术过程对增强现实系统进行了设计，主要分成四个部分，包括虚拟颅骨模型的三维重建、增强现实系统的空间定位、真实手术场景的重构和虚实融合显示。采用基于Matlab工具箱的摄像机标定方法，完成了系统的空间定位，为增强现实融合部分做了前续工作。 2.设计了易于识别、提取的蓝色外加标志点，利用目标窗口对颅骨定位剪切，提取蓝色标志点，计算标志点质心坐标并对质心标号，利用基于Laplacian矩阵的图像匹配方法对左右图像进行匹配，最后利用加权图像融合方法和像素灰度值极大（小）图像融合法完成左右颅骨图像外表面二维融合。 3.研究了虚实配准的原理、方法，提出了一种融合图像显示方法。利用三维扫描颅骨的数据重建虚拟颅骨模型，采用了对多角度图片进行三维场景恢复的方法得到真实手术场景，，最后把虚拟的颅骨模型导入到真实的手术场景空间去，显示在计算机上。本文在搭建的实验平台的基础上，对该系统进行了实验研究。首先进行双目摄像机标定，经过实验分析，系统的标定误差在0.3个像素以内，达到了亚像素级的精度；然后对左右摄像机采集的图像进行Laplacian匹配和加权、像素灰度值极大（小）图像融合；最后将重建的虚拟模型导入到真实手术场景中并通过计算机屏幕进行显示。本文研究的阶段性成果为后续动态可视化研究打下了基础。
[Abstract]:The image fusion technology based on augmented reality, which is applied in the surgical navigation system, has changed the traditional operation method. It can not only reduce the pain of patients, but also reduce the risk of surgery, and make the preoperative planning accurately. Also has the very high application value in the medical treatment training aspect. The image fusion based on augmented reality (AR) overlay the virtual lesion on the real surgical site, which provides more support for the doctor's diagnosis. In this paper, the fusion of two dimensional images and three dimensional images are studied. Through the fusion of the external surface of the site to be operated on and the three dimensional fusion of the internal physiological information, the amount of information of the focus is increased. The specific research contents and results are as follows: 1. In this paper, the augmented reality system is designed according to the actual operation process, which is divided into four parts, including the 3D reconstruction of virtual skull model, the spatial location of augmented reality system, the reconstruction of real operation scene and the display of virtual reality fusion. The camera calibration method based on Matlab toolbox is used to complete the spatial positioning of the system. The blue additional mark points, which are easy to recognize and extract, are designed. The target window is used to locate and cut the skull, the blue mark points are extracted, the centroid coordinates of the mark points are calculated, and the centroid of mass is labeled. The image matching method based on Laplacian matrix is used to match the left and right images, and the weighted image fusion method and the maximum (small) pixel gray value fusion method are used to complete the 2D fusion of the left and right skull images. The principle and method of virtual real registration are studied, and a fusion image display method is proposed. The virtual skull model is reconstructed by using the data of 3D scanning skull, and the real operation scene is obtained by using the method of 3D scene restoration of multi-angle images. Finally, the virtual skull model is imported into the real operation scene space. Display on the computer. On the basis of the experimental platform, the system is studied experimentally in this paper. The calibration error of the system is within 0.3 pixels, and the accuracy of sub-pixel level is achieved, then the images collected by the left and right cameras are matched and weighted by Laplacian. Finally, the reconstructed virtual model is imported into the real operation scene and displayed on the computer screen. The results of this paper lay a foundation for further dynamic visualization research.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R616;TP242;TP391.41

【参考文献】