锥束CT感兴趣区域图像重建研究

针对锥束CT感兴趣区域扫描中存在的截断投影数据图像重建问题,提出用基于迭代的代数重建(ART)算法进行重建。锥束ART算法的缺点是计算量大、重建速度慢。为了提高该算法的重建速度,提出了一种基于多核平台的快速并行图像重建方法。首先将三维重建区域等分为上下两块,相应地,探测器平面也分为上下两部分;然后通过双线性插值计算虚拟探测器投影数据;最后通过多线程技术在多核平台上实现了ART算法的并行重建,在保持较高重建精度的同时取得了约两倍的重建加速比。在此基础上,通过仿真实验对3DShepp-Logan模型不同感兴趣区域进行了重建,实验结果表明,ART算法用于感兴趣区域图像重建是可行的。

Image reconstruction of cone-beam CT for region of interest

The projection data is often truncated during scanning within the region of interest ROI by cone-beam CT, which results in incompleteness of projection data. Aiming at the problem of image reconstruction from incomplete projection data, this paper proposed ART algorithm based on iteration to reconstruct for ROI. The shortage of ART was its large time of computations and slow reconstruction speed. To improve the reconstruction speed of ART, It proposed a fast parallel image reconstruction method based on multi-core. Firstly, it divided the 3D reconstruction region equally into two blocks by centre plane, and the detector plane too. Then, it calculated the projection data of virtual detector by bilinear interpolation. Finally, imlemented the ART algorithm parallel using multi-thread technique on multi-core platform, and improved the reconstruction speed about 2 times with high precision. On this basis, using ART to reconstruct different ROI of 3D Shepp-Logan phantom by simulation, experimental results show that it is feasible to reconstruct ROI with ART.