CASIA OpenIR  > 毕业生  > 硕士学位论文
基于CUDA架构的CT在线重建与可视化研究
Alternative Titlefast CUDA-based on-line CT reconstruction and volumn rendering
李勇保
Subtype工程硕士
Thesis Advisor杨鑫
2013-05-19
Degree Grantor中国科学院大学
Place of Conferral中国科学院自动化研究所
Degree Discipline计算机技术
KeywordCt重建 Fdk加速 Cuda 体绘制 在线重建 Ct reconstruCtion Fdk Cuda Volumn Rendering On-line Reconstruction
Abstract自1895年德国科学家伦琴发现X射线以来,医学诊断已经发生了翻天覆地的变化。计算机断层成像(CT)作为一种无损的成像方式,使得医疗诊断有了革命性的发展,它也成为了现代放射科的基础设备。三维重建和重建结果的可视化是CT成像最重要的两步。准确快速的重建提供了物体的吸收系数分布,让我们获得了丰富的结构信息。可视化技术直观清晰的展示了这种信息,从而让临床诊断更加可靠。 随着X射线探测器的快速发展,投影采集速度越来越快,像素的个数和精度也在不断增加,这对CT的重建速度和可视化速度提出了更高的要求。GPU的发展以及CUDA架构的提出,为CT加速重建尤其是FDK锥束重建和可视化加速提供了新的思路。与传统的加速方式相比,CUDA编程方便,入门门槛较低,且花费低廉。 本文主要针对在线FDK锥束重建进行了研究。在线重建可以让我们快速清楚的观察到投影数据逐步合成三维物体的过程,这种实时反馈在临床和工业上有着重要意义,也是4D CT可视化的一种途径。本文主要涉及的内容有:滚筒式Micro-CT系统的平台搭建,基于CUDA的FDK锥束重建加速和体绘制加速,以及在线重建的设计和实现。主要贡献包括: 1、滚筒式Micro-CT小动物成像系统的研究。设计并实现了一套高分辨率的Micro-CT成像系统,内容涵盖了主要部件的选型和设计,并通过对采集图像进行预处理降低了图像的噪声和重建伪影。最终,系统可以方便的调整成像范围和重建精度,通过多个试验结果,验证了系统的稳定性和有效性。 2、利用CUDA架构对FDK锥束重建算法进行了加速,并给出了详细的实 现方法。其包括设计线程和BLOCK维数实现全局存储器的合并访问优化,运用纹理存储器加速插值计算,调用共享存储器加速滤波过程以及投影幅数选择等等。通过大量的实验分别验证了每个方法的效果,并和国际主流算法进行了对比。 3、利用CUDA架构实现光线投射算法的加速,包括最大密度算法,等面值绘制算法以及合成体绘制算法的实现等等。对这三种算法进行了对比,并给出了可视化结果。最后结合实验室搭建的小动物Micro-CT系统,设计并优化了Micro-CT的在线重建过程。
Other AbstractSignificant changes have taken place in medical diagnosis since German scientistRoentgen discovered X beam in 1895. Computed Tomography(CT) is a non-destructive imaging method and the foundation equipment of modern radiology. Reconstruction and the visualization of reconstruction volumes are two main steps in CT imaging. Fast and accurate reconstruction shows three-dimensional absorption coefficient distribution of an object, and thus we get a wealth of structural information. Intuitive and clear visualization techniques show this structural information so that clinical diagnosis becomes more reliable. With the rapid development of X beam detector, the number of detector pixels and the resolution are increasing. At the same time, the acquisition speed of projected images becomes faster which puts forward higher requirements for reconstruction and visualization speed. The development of GPU and CUDA architecture provide a new way for CT accelerated reconstruction and visualization, especially for FDK cone-beam reconstruction. Compared with the traditional acceleration way, CUDA programming has more convenience, lower barriers to entry and cost less. We mainly study online FDK cone-beam reconstruction in the paper. Online reconstruction allows us to clearly observe the gradual synthesis process of three-dimensional volume from projection data. This real-time feedback has a great significance in industry and clinical applications and is also a way of 4D CT visualization. This article is primarily related to the build of drum Micro-CT imaging platform, CUDA-based FDK accelerated reconstruction and visualization and the implementation of online CT reconstruction. The main work and contributions are as below: 1. Implementation on drum Micro-CT imaging platform. This part involves the selection and design of the main components, and reducing image noise and reconstruction artifacts through preprocessing on projection images. The final system can easily adjust imaging range and reconstruction accuracy. Via a plurality of test results, we verify the stability and effectiveness of this system. 2. CUDA-based FDK cone-beam reconstruction acceleration. In this part, we give a detailed acceleration method including using constant memory to store sine values and cosine values, realizing consolidated global memory access optimization by designing thread and block dimensions, accelerating interpolation with texture memory and filtering process with shared memory....
shelfnumXWLW1928
Other Identifier2010E8014669005
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/7683
Collection毕业生_硕士学位论文
Recommended Citation
GB/T 7714
李勇保. 基于CUDA架构的CT在线重建与可视化研究[D]. 中国科学院自动化研究所. 中国科学院大学,2013.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_2010E801466900(13962KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[李勇保]'s Articles
Baidu academic
Similar articles in Baidu academic
[李勇保]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[李勇保]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.