CASIA OpenIR  > 毕业生  > 博士学位论文
海量医学影像数据可视化算法及集成化算法平台的研究
其他题名Research on Visualization Algorithms and Integrated Algorithm Platform for Out-of-Core Medical Data
薛健
学位类型工学博士
导师田捷
2007-06-20
学位授予单位中国科学院研究生院
学位授予地点中国科学院自动化研究所
学位专业计算机应用技术
关键词医学影像 海量医学影像数据处理 三维交互 三维可视化 体绘制 算法平台 Medical Imaging Out-of-core Medical Data Processing 3d Interaction 3d Visualization Volume Rendering Algorithm Toolkit
摘要随着医学影像设备成像手段和硬件技术的不断进步,医学影像的空间分辨率越来越高,随之而来的是采集影像的数据量急速上升,这给已有的医学影像处理算法带来了严峻的挑战。从算法研发的基础平台层面看,目前国际上主流的算法平台(如VTK、ITK 等)并没有在基础层面建立相应的算法框架提供对海量数据处理的支持,使得这些算法平台在面对海量数据时失去了其作为算法研发支撑平台所应起到的作用。 本文的工作即针对上述问题而展开,其中包含两个方面:一是针对海量(Out-of-Core)医学影像数据处理和可视化的基础算法平台的研究和实现;二是对海量医学影像数据可视化算法的研究。主要工作包括下面四个部分: 1. 提出了一套面向海量医学影像数据处理及可视化的统一算法框架设计,包括整体框架、访问接口和底层数据结构的设计,并将其在已有的算法平台基础上予以实现,完成了海量医学影像数据处理与分析算法平台MITK(Medical Imaging ToolKit)2.0 的设计和实现。 2. 提出了一种基于3D Widgets,面向海量医学影像可视化的三维交互框架设计,并在海量医学影像数据处理与分析算法平台中予以实现,一方面为海量医学影像数据的可视化算法提供了一个相对独立、简单易用并且便于扩展的三维交互支持框架,另一方面在可视化算法的定性显示与实际应用中所需要的定量分析之间建立起了沟通的桥梁。 3. 提出了一种基于半自适应分块和硬件加速的海量医学影像数据光线投射(Volume Ray Casting)直接体绘制算法。传统光线投射体绘制方法是从某一特定视点经过成像平面发射光线穿过体数据场,沿着这些光线在体数据场中采样计算,并按从前往后的顺序合成最终投影图像。这种方法在计算采样点的光属性时,不管采用何种插值方式,都需要对原始体数据做随机访问,这样在处理海量数据时必然带来过多的内外存数据交换而导致严重的系统颠簸,从而使绘制效率大大降低。而本文提出的算法采用一种半自适应分块的策略对原始数据进行合理的分块,尽量分离不需绘制的背景部分,用BSP树来组织所有子数据块,从而能够保证按正确的顺序从前往后绘制所有子数据块,并且所有数据只导入内存一遍,避免了对原始数据的随机访问,同时,通过在绘制时忽略完全是背景的子数据块以及用GPU来加速每个子数据块的光线投射计算,大大提高了海量数据光线投射体绘制的效率。 4. 提出了一种基于三维纹理加速的海量医学影像数据直接体绘制算法。本算法通过对海量原始数据进行合理分块,使每块数据能够被装入内存和显卡显存单独处理,从而能够对海量医学影像数据进行体绘制。同时,通过使用三维纹理硬件和改进纹理多边形生成方法等措施优化分块绘制流程,大大加速整个绘制过程。
其他摘要With the rapid development of medical imaging devices and related techniques, the resolution of the acquired images are getting higher and higher, which leads to very large (out-of-core) medical data sets and challenges the conventional medical image processing and analyzing algorithms. Furthermore, most mainstream algorithm toolkits and application systems do not provide the supports for out-of-core data processing, which makes these toolkits loose the function of supporting platform for the algorithm research and development in the out-of-core situation and restrict their application for very large medical data sets which appear more and more these years. The research works of this dissertation just aim at above problems and include following parts: 1. Proposed a set of algorithm frameworks for out-of-core medical data processing and visualization including the design of the total framework, access interfaces and underlying data structures, implemented them in our former basic algorithm platform, and completed the design and implementation of the new out-of-core medical data processing and analyzing algorithm toolkit MITK (Medical Imaging ToolKit) 2.0. 2. Proposed a 3D human computer interaction framework based on 3D widgets for out-of-core medical data visualization and implemented it in the new out-of-core medical data processing and analyzing algorithm platform. It not only provides a comparatively independent, simple-to-use and extensible 3D interaction framework for the out-of-core medical data visualization algorithms, but also sets up a communicating bridge between the qualitative display of the visualization algorithms and the quantitative analysis needed by actual applications. 3. Proposed an efficient out-of-core volume ray casting algorithm for the visualization of very large (out-of-core) medical data sets based on semi-adaptive partitioning and graphics hardware acceleration. It applied a semi-adaptive partitioning strategy to split the original volume data into sub-blocks with appropriate size and separate the background region as exact as possible. BSP tree is also used to organize all the sub-blocks so as to ensure that all the sub-blocks are rendered in front-to-back sequence and avoid random accesses to the original data. Furthermore, when rendering the BSP tree, the empty sub-blocks are skipped and GPU is used to accelerate ray casting computation in each nonempty sub-block. 4. Proposed a 3D texture-based out-of-core volume rendering algorithm for accelerating the visualization of very large medical data sets. In order to render out-of-core volume data sets with limited 3D texture buffer, this algorithm divides the volume data into small blocks, applies 3D texture hardware and an improved method for generating texture polygons to optimize the rendering process of each block, so as to accelerate the total rendering procedure enormously.
馆藏号XWLW1127
其他标识符200418014628057
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/6031
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
薛健. 海量医学影像数据可视化算法及集成化算法平台的研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2007.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
CASIA_20041801462805(10273KB) 暂不开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[薛健]的文章
百度学术
百度学术中相似的文章
[薛健]的文章
必应学术
必应学术中相似的文章
[薛健]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。