虚拟内窥镜技术研究

【作者】 张国华；

【导师】 杨兴强；

【作者基本信息】 山东大学 ， 计算机应用技术， 2005， 硕士

【摘要】 虚拟内窥镜技术(Virtual Endoscopy)是虚拟现实技术在现代医学中的应用,是利用医学影像作为原始数据,融合图像处理、计算机图形学、科学计算可视化、虚拟现实技术,模拟传统光学内窥镜的一种技术。近年来,随着高分辨率医学图像(CT、MRI等)的出现和虚拟现实技术的不断完善、提高,虚拟内窥镜技术得到了迅速发展,成为虚拟现实技术走向实际应用最为活跃的研究领域之一。虚拟内窥镜的研究旨在为医生提供诊断依据,还可应用于辅助诊断、手术规划、实现手术的精确定位和医务人员的培训等。它克服了传统光学内窥镜需把内窥镜体插入人体内的缺点,是一种完全无接触式的检查方法,在医疗领域中得到了越来越广泛的应用。本文针对虚拟内镜系统的关键技术和实现方法进行了深入研究。 首先,文章分析了在CT、磁共振、医院PACS系统中应用广泛的医学图像数据DICOM文件格式,进而实现了对DICOM文件的正确解读,以提取三维重建必需的信息,如病人的检查部位、图像的窗宽窗位以及切片厚度等。对于三维重建,文章简要介绍了几种常用的表面绘制算法,并对Marching Cube算法的实现流程作了说明。 其次,文章设计并实现了一种交互式的半自动图像分割方法——曲线切割。对于重建后的图像,感兴趣区域有可能被遮挡,这时一般需要手动切除部分组织。本系统中的切割,只需要用户拖动鼠标形成一条封闭曲线来标识出要切除的区域。直接判断每个体素的投影在曲线内外,非常耗时,针对这一点,本文提出了一种基于区域增长的快速有效的切割方法,并在不增加占用内存的情况下实现了切割操作的撤销和恢复。 再次,文章详细介绍了给定起止点的漫游实现的具体过程和方法。漫游实现的关键一步是空腔组织的中心路径抽取,即骨架化,文章对已有的骨架化算法进行了分析和改进,提高了系统的运行效率。 最后,由于直接沿骨架化算法的结果进行漫游容易出现图像跳跃现象,文章提出了用NURBS曲线进行平滑路径的方法,并进一步研究了权因子变化对NURBS曲线形状的影响规律。以NURBS曲线的齐次坐标表示为基础从几何角度证明了一个权因子和两个权因子的变化对曲线形状影响的规律,从力学角度揭示了权因子的物理意义,为更有效地利用权因子提供的额外自由度来调整曲线形状提供了理更多还原

【Abstract】 Virtual Endoscopy (VE) is an application of Virtual Reality in modern medicine. It is a technique incorporating the research about Computer Graphics, Image Processing, Scientific Computing Visualization, Virtual Reality, etc. VE simulates optics endoscopy by use of medicine images. With the advent of high resolution medical images (such as CT, MR I etc. )and the progress of Virtual Reality technique , VE is developing rapidly . It has been one of the most flourished fields during the application of VR to practice. The aim of VE system research is to provide trusty means for medical diagnoses, and to spread the application to the fields such as assistant diagnoses, surgery planning and training. VE is a non-touched medical diagnoses method and is becoming more and more widely used in many medical fields since it avoids the risk of invading body when the actual optical endoscopy is used. Some key techniques and realizations of VE are systematically investigated in this dissertation.Firstly, the medical image of DICOM is analyzed. DICOM file format is widely applied in CT, MRI and PACS systems. When DICOM files is read correctly, the information needed for 3D reconstruction, such as the patient’s part examined, the window width, the window center and the slice thickness of the image and so on, will be extracted. Then, some kind of 3D reconstruction based on surface rendering algorithms are introduced, and the realization process of Marching Cube algorithm is addressed briefly.Secondly, an interactive and semi-automatic image segment method, named curved-cutting, is designed in this system. It is necessary to cut off part of tissues when the interested region is hided as a 3D object is displayed. In our system, it is enough when the user draw a closed curve by moving the mouse. It is very time-consuming to judge the projectionof every voxel is in or out of the curve directly. So a fast method based on region-growing is put forward and realized. And besides ,the segment operation of undo/redo is realized without extra memory.Thirdly, the process and method of the realization of fly-through are discussed in detail. The key step of getting the hollow tissue’s center path , also named skeletonization , is another time-consuming work. Then one of the skeletonization algorithms available is analyzed and improved.Finally, practical path-smoothing algorithm based on NURBS is given since image jumping often occurs when flying along the result of skeletonization. Furthermore, with the representation of homogeneous coordinates, the shape modification based on weight is discussed from the point of geometry. Physical meaning of weight is uncovered and then theoretic evidence is provided to modify curve shape by using weight effectively.更多还原