【作者】 李彬；

【导师】 毛宗源； 田联房；

【作者基本信息】 华南理工大学 ， 控制理论与控制工程， 2007， 博士

【摘要】 在进行三维适形放射治疗之前,需要运用三维医学可视化技术,获取有关病灶区的三维视觉信息(病灶的体积、空间定位、病灶与重要器官间的相对位置等),并确定靶区,以供治疗模拟与方案评估,从而实现精确地计划。论文系统地分析了三维可视化技术的特点,围绕着“普通硬件条件下如何高效率地重建质量较好,并能满足三维适形放射治疗计划要求的三维医学图像”的研究重点,有针对性地提出一些技术改进和创新算法,并综合这些技术开发了用于三维适形放射治疗计划的三维可视化系统,目前正在临床测试中。首先,根据三维适形放射治疗计划的三维可视化目标,研究了基于模块化设计的医学体可视化算法,并详细介绍了各个模块的实现过程:利用基于混合型链结区域增长与聚类分析的分割方法,提取需要重建的组织;应用最大互信息法实现解剖图像CT与功能图像PET的配准;应用基于小波变换的融合方法,以区域标准差与区域能量相结合的融合规则实现解剖图像CT与功能图像PET的融合。采用基于梯度与等值面的光线吸收与发射模型以及改进的Shear-Warp算法,实现融合后序列图像的体绘制。该部分为后续研究奠定了基础。其次,针对寻求最佳传递函数的问题,研究了一种改进的体绘制算法(Improved Shear-Warp Algorithm,ISWA)。在传递函数预设定时,基于边界模型,结合对象分析和交互式操作,可以方便地找到最优传递函数值;在重采样过程中,结合Shear-Warp算法的特点,对传递函数值进行修正,以保证重建后图像效果。针对透视投影的Shear-Warp算法的不足:1)切片间采样的混淆现象,使得绘制结果通常难以保证医学图像质量要求;2)在不透明度传递函数和观察方向需要频繁变换的医学可视化交互式操作中,重建速度有待进一步提高,提出了一种改进的透视体绘制算法(Improved Perspective Volume Rendering Algorithm,IPVRA)。利用不透明度传递函数的相关性,以及min-max Octree数据结构,实现对体数据的快速分类编码,并在此基础上,快速再建游程编码数据结构,以满足在不透明度传递函数和观察角度频繁改变时,也能实现快速重建;基于透视Shear-Warp算法,采用结合光照模型的改进预合成体绘制技术,消除因欠采样而产生的混淆现象,提高了重建的效果。再次,为了提高多分辨率体绘制成像的质量与效率,采用将体绘制与数据压缩、解压进行有机的结合的思路,基于小波变换和Shear-Warp算法,研究了一种快速多分辨率体绘制方法(Fast Multi-resolution Volume Rendering Method,FMVRM)。该方法先基于小波变换对体数据进行压缩;接着结合所选的分辨率,利用不透明度传递函数引导数据的解压;最后,基于Shear-Warp算法,利用基于子数据块的游程编码数据结构、不透明度传递函数的相关性和结合光照模型的预合成体绘制技术,实现快速而高质量的多分辨率体绘制。为了高效率地重建三维医学图像,并能充分显示内部隐含分界面及内部细节的详细信息,提出了一种能在普通硬件条件下实现的,用于半透明体绘制的改进的光学模型(Improved Volume Rendering Optical Model,IVROM)。该模型在光线吸收与发射模型的基础上,考虑了阴影、直接散射与间接散射等因素。文中详细介绍了结合该模型与Shear-Warp的半透明体绘制方法的实现。为了提高医学体数据与几何模型,以及不同分辨率体数据之间的混合体绘制图像的视觉效果和效率,研究了一个基于Shear-Warp算法的普适混合体绘制方法。算法首先利用Z-Buffer算法与游程编码RLE数据结构对各类重建数据进行预处理;接着,在合成中间图像的过程中,采用二重重采样方法,依据精度要求选择重采样方式进行重采样,并用即时插值中间切片的重采样方法解决采样过疏问题。最后,用OpenGL纹理技术绘制最终图像。最后,综合上述的研究成果,根据三维适形放射治疗计划的需要,开发了用于三维适形放射治疗计划的三维可视化系统。实验证明,运用该三维可视化系统,可以获取有关病灶区的三维视觉信息并确定靶区,以供治疗模拟与方案评估。更多还原

【Abstract】 When 3D conformal radiotherapy (3D CRT) is employed for tumor treatment, the 3D visual information about tumor, including the volume and the positioning of the tumor, and as well as the relative position between the tumor and its adjacent tissues, should be obtained using 3D visualization techniques, so that the target is determined for virtual radiotherapy and program evaluation, thus accurate planning can be realized.In the thesis, the characteristics of the visual techniques are systematically analyzed, and the reconstruction of 3D medical image with high quality that can be applied for 3D CRT planning is emphasized, which can be run on a general-purpose PC with high efficiency. On this purpose, some improvements for current used methods and innovative algorithms are proposed, and a 3D visual system is developed for 3D CRT by using those proposed techniques.Firstly, based on the aim of 3D visualization for 3D CRT planning, volume visualization methods for medicine designed in modularization are studied, they are: the segmentation module, which is based on hybrid-link region growing method and cluster analysis algorithm for the purpose of the segment of the tissue to be reconstructed; the registration module, which is based on maximum mutual information method to realize the registration of anatomical image (CT) and functional image (PET); the fusion module, which is based on wavelet theory with fusion rule of combining the local standard deviation and energy to realize the fusion of PET and CT images; the volume visualization module, which is based on the improved Shear-Warp algorithm on the basis of ray absorption and emission model to realize volume rending of the series of fused images. Those works establish the foundation for further research.Secondly, for the solution of optimum transfer function, an improved Shear-Warp algorithm(ISWA) is proposed. An optimal transfer function can be conveniently found by combining boundary model and interactive operation when pre-specifing transfer function. In re-sampling process, the parameters of the transfer function is modified to guarantee the reconstruction effect of the image.Thirdly, as to the shortcoming of perspective Shear-Warp algorithm, an improved perspective volume rending algorithm (IPVRA) is developed. In which rapid classification coding for volume data is implemented by making use of the correlation of opacity transfer function and min-max Octree data structure. On this basis, the new run-length encoded(RLE) data structure can be established rapidly so that fast reconstruction can be done when opacity transfer function and view angle change. Moreover, the aliasing artifacts resulted from under-sampling can be eliminated by pre-integrated volume rendering technique incorporated the shading calculation of Phong lighting model, thus improving the reconstruction effect.Fourthly, in order to enhance the quality and efficiency of multi-resolution volume rendering method, a fast multi-resolution volume rending method (FMVRM) is proposed by combining volume rendering and data compress and decompress technique. The technique is as follows: in the beginning, data compress is carried using wavelet transformation, then according to the chosen resolution, data decompression guided by opacity transfer function (OTF) is implemented. After that, based on Shear-Warp algorithm, and by making use of block-based run-length encoded data structure,the correlation of opacity transfer function and pre-integrated volume rendering technique incorporated the shading calculation of Phong lighting model, a fast volume rending with high quality can be realized.In order to improve the reconstruction efficiency of medical image, and sufficiently display the internal hidden interfaces and the details of internal information, an improved volume rendering optical model(IVROM) for translucent volume rendering is developed, which can be rapidly executed on a general-purpose PC. In the model, the lighting absorption and emission model are employed, moreover, the factors of volumetric shadowing, direct and indirect emission are also considered. In addition, the realization of translucent volume rendering method combining the IVROM and Shear-Warp algorithm is presented in detail. In order to enhance the visual effect and reconstruction efficiency of hybrid volume rendering of polygonal models and different-resolution medical volume data running on a general-purpose PC, a hybrid volume rendering algorithm is proposed based on Shear-Warp algorithm. In which, various data for reconstruction are first pre-processed by employing Z-Buffer algorithm and RLE data structure, then a dual re-sampling method is chosen according to accuracy requirement for compositing intermediate image and the intermediate slice interpolation method is employed to solve the problem of over-sparse of re-sampling. Finally, the reconstructed image is rendered using OpenGL texture-mapping technique.Finally, a 3D visualization system for radiotherapy treatment planning is established based on the above technique. Experiments demonstrate that the 3D visual information of the tumor and target volume can be precisely determined using the proposed strategies, thus providing a good means for radiotherapy treatment and program evaluation.更多还原