三维医学数据体绘制的传输函数设计方法研究

发布时间：2017-12-27 20:04

本文关键词：三维医学数据体绘制的传输函数设计方法研究　出处：《中国科学技术大学》2016年博士论文　论文类型：学位论文

【摘要】：随着现代医学影像学和计算机技术的高速发展,三维医学数据可视化在医学影像的诊断与治疗以及生物医学研究中起到越来越关键的作用。其主要作用是帮助临床医生或研究人员直观的看到并了解生物体内部的情况,尤其是他们感兴趣的组织或结构的相互连接关系和几何形状大小。相比于传统的二维切片显示,三维医学数据的可视化为用户提供了更加逼真的显示手段和定量的分析工具,便于用户从多角度、多层次的观察分析医学影像,在辅助诊断、引导治疗等方面都有着广泛的应用。三维医学数据可视化常用的一种方法是直接体绘制,传输函数是直接体绘制中最重要的步骤之一,它把三维医学影像数据中每个体素的数值特征映射成为光学特征,如颜色和不透明度,实现具有真实感的三维图像绘制。传输函数设计的好坏决定了可视化的质量的高低。本文的主旨在于研究高效的传输函数设计方法和体绘制交互方法,其目的都是避免复杂的交互式传输函数设计,提升医学体绘制的效率。经典的基于边界的传输函数设计,如f-|%絝|二维统计直方图和LH二维统计直方图,能够快速的聚类和提取不同的边界。然而,这两种方法在噪声干扰下,直方图中边界的聚类变得模糊,这导致边界提取不完整和用户交互复杂。由于各个角度体数据的投影不同,需要调节传输函数来使得感兴趣的组织或结构得到突出显示。然而,体绘制参数的调节对于没有先验知识的普通用户来说,是一个反复尝试而且耗时的过程。针对上述问题,本文从基于边界的传输函数设计和体绘制参数的调节两个方面,对上述算法进行改进,提升了边界提取的可靠性和用户交互的高效性。1、针对CT图像,本文提出了一种基于动态的统计直方图的传输函数设计方法。一方面,我们建立了一个通用的受噪声干扰的边界模型：另一方面,在此基础之上,我们提出了一个新的基于边界中值(M)、边界高度(△h)和梯度(|%絝|)的多维传输函数设计方法。和传统的统计直方图呈现所有的边界信息不同,我们提出的动态直方图方法每次只显示一种边界。另外,还提出了一个简单的边界迭代提取策略：首先把所有边界按照边界高度由高到低排序,用户控制△h的值由大到小直到有一个竖条出现在M-|%絝|直方图中,边界会被依次提取直到△h变为O。得益于这个迭代提取策略,边界间的误分和受噪声干扰得以降低；此外,由于区域消除和区域扩张的进一步运用,绘制质量得以提高。2、本文亦提出了一种所选即所得(what material you pick is what boundary you see)边界体绘制传输函数设计方法。不同于传统的直方图的边界传输函数设计方法是利用数学统计聚类实现的,我们首先提出了以用户为中心的边界提取的准则和基于这个准则改进的边界模型,然后我们提出了一个所选即所得的边界传输函数设计方法。一方面,用户可以通过直接选择感兴趣的组织或结构来传递语义信息；另一方面,我们采用三维Canny边缘检测算子来确保边界位置的可靠性。最后,我们建立了一个点到物体(point-to-material)的距离度量来确保边界的准确性和完备性。整个基于边界的传输函数设计方法都是直接在三维图像上进行操作,设计过程具有很好地直观性、实时性和精确度。3、直接体绘制把三维物体压缩成二维图像会造成深度信息的损失,这种损失主要表现为物体的遮挡。为了快速调节传输函数参数使得感兴趣的组织或结构能在各个角度都能有清晰突出的显示,本文首先提出可视比率(visibility ratio)和遮挡向量(occlusion vector)两个定量描述遮挡的概念,这两个概念都是直接来源于离散体绘制算法的推导。其次我们提出了一个交互式的线性反馈机制来调节传输函数的参数。整个反馈机制与经典的a颜色混合非常类似,用户可以定量的了解遮挡和通过直觉调整参数。另外,传输函数参数调节不涉及到具体的传输函数设计方式,可以把传输函数参数调节方法集成到任意的以数据为中心的直接体绘制系统中作为一个重要的补充。经过大量的实验证明,本文提出的传输函数设计和调节方法能较好的检测和提取边界,减少用户的交互,提高医学数据可视化的效率。
[Abstract]:With the rapid development of modern medical imaging and computer technology, 3D medical data visualization plays a more and more important role in medical imaging diagnosis and treatment as well as biomedical research. Its main role is to help clinicians or researchers intuitively see and understand the internal situation of organisms, especially the interconnections and geometry size of organizations or structures that they are interested in. Compared with the traditional 2D slice display, the visualization of 3D medical data display and analysis tools provides more realistic and quantitative for the user, is convenient for users to observe from the multi angle, multi-level analysis of medical images, are widely used in the diagnosis and treatment guide. A commonly used method of 3D visualization of medical data is direct volume rendering, the transfer function is one of the most important steps in direct volume rendering, the numerical characteristics of 3D medical image data and maps each voxel as optical characteristics, such as color and opacity, 3D image with photorealistic rendering. The quality of the visual quality is determined by the design of the transmission function. The purpose of this paper is to study efficient transmission function design method and volume rendering interaction method. The purpose is to avoid complex interactive transfer function design and improve the efficiency of medical volume rendering. The design of transfer function based on the classical boundary, such as f-|% Ku | two-dimensional statistical histogram and LH two-dimensional histogram, clustering and extraction can fast different boundary. However, under the noise interference, the clustering of the boundary in the histogram becomes blurred, which leads to the incomplete boundary extraction and the complexity of the user interaction. Because the projection of each angle body is different, it is necessary to adjust the transmission function to make the organization or structure of interest be highlighted. However, the adjustment of volume rendering parameters is a trial and time-consuming process for ordinary users without prior knowledge. To solve these problems, we improve the above algorithms from two aspects, which are boundary based transmission function design and parameter adjustment of volume rendering, which improves the reliability of boundary extraction and the efficiency of user interaction. 1. In view of the CT image, this paper presents a design method of the transmission function based on the dynamic statistical histogram. On the one hand, we establish a general model of the noise in the boundary. On the other hand, on this basis, we propose a new boundary based on median (M), boundary height (delta H) and gradient (|% Ku |) the multi-dimensional transfer function design method. And the traditional statistical histogram shows that all the boundary information is different. The dynamic histogram method proposed by us shows only one boundary each time. In addition, also proposed the strategy of extracting a simple iterative boundary: first of all according to the boundary boundary from low to high ranking, the user to control the delta H values from large to small until there is a bar in M-|% Ku | histogram, the boundary will be extracted in turn until the H is O. Thanks to the iterative extraction strategy, the erroneous points between boundaries and noise interference are reduced. Moreover, due to the further application of region elimination and region expansion, the rendering quality is improved. 2. This paper also presents a design method for the transfer function of the what material you pick is what boundary you see) boundary volume rendering. The boundary transfer function design method is different from the traditional histogram is the use of mathematical statistical clustering, we first propose a user centric boundary extraction criteria and improvement based on this criterion boundary model, then we propose a design method that is selected from the boundary of the transfer function. On the one hand, users can transmit semantic information directly by selecting organizations or structures of interest. On the other hand, we use three-dimensional Canny edge detection operators to ensure the reliability of boundary locations. Finally, we set up a distance measure of the point to object (point-to-material) to ensure the accuracy and completeness of the boundary. The transfer function design method based on boundary are operated directly on the three-dimensional image, the design process has very good direct view, real-time and accuracy. 3. The loss of depth information is caused by the compression of three-dimensional objects into two-dimensional images by direct volume rendering, which is mainly represented by the occlusion of the object. In order to quickly adjust the transfer function parameters to make the interests of the organization or structure can display clear and outstanding in all angles, this paper put forward visibility ratio (visibility ratio) and occlusion (occlusion vector) vector concept two quantitative description of occlusion, the two concepts are derived directly from the discrete rendering algorithm. Secondly, we propose an interactive linear feedback mechanism to adjust the parameters of the transmission function. The whole feedback mechanism is very similar to the classic a color mixing, and the user can quantify the occlusion and adjust the parameters by intuition. In addition, transfer function parameter adjustment does not relate to specific transmission function design. It can integrate transmission function parameter adjustment into any data centric direct volume rendering system as an important supplement. After a lot of experiments, it is proved that the transmission function design and adjustment method proposed in this paper can better detect and extract boundaries, reduce user interaction and improve the visualization efficiency of medical data.
【学位授予单位】：中国科学技术大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.41;R318

【相似文献】