有序子集最小二乘OS-LS图像重建迭代算法  

Study on Ordered Subsets-least Square Reconstruction of Image

作　　者：刘力[1] 印胤[1] 单保慈[1] 

机构地区：[1]中国科学院高能物理研究所,北京100049

出　　处：《中国图象图形学报（A辑）》2005年第5期628-632,共5页Journal of Image and Graphics

基　　金：国家自然科学基金项目(90209030)

摘　　要：为推导一种新的快速图像迭代重建方法,首先将有序子集(orderedsubsets,OS)技术应用到最小二乘图像重建迭代算法(leastsquarereconstruction,LS);然后对仿真Phantom模型数据和实际医用正电子发射断层成像仪(PET)数据进行重建,并研究了在不同子集划分下的重建结果,同时分析比较了不同子集的选取对OSLS重建图像质量以及重建收敛速度的影响。重建结果表明,这种基于有序子集的最小二乘图像重建迭代算法(OSLS)具有较高的重建图像质量和较短的计算时间,相对于传统LS算法的重建,OSLS的收敛速度加速了约L倍(L为子集个数),其重建图像质量也好于传统的滤波反投影(FBP)方法的重建,可应用在PET图像重建中。In order to construct a new practical and fast iterative image reconstruction method, the ordered subsets(OS) technique is combined with the least square(LS) reconstruction of images in medical tomography. Reconstruction of simulated data and real positron emission tomography(PET) data shows that so accelerated OS-LS iterative image reconstruction method has a rapid convergent speed and higher spatial resolution. The reconstructed image quality and convergence speed by different subsets order are studied. As compared to the traditional LS reconstruction, OS-LS is L times faster, where L is the number of subsets, and the reconstructed images by OS-LS are better than the conventional FBP (filtered back-projection) as well. The conclusion is that the so proposed OS-LS reconstruction method can be used in real PET image reconstruction.

关 键 词：图像重建 迭代算法 最小二乘 子集 OS 有序 PHANTOM 收敛速度 图像质量 正电子发射 滤波反投影 重建方法 技术应用 模型数据 分析比较 计算时间 LS算法 成像仪 PET 传统 仿真 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] O177.91[自动化与计算机技术—计算机科学与技术]