中国组织工程研究

中国组织工程研究 ›› 2011, Vol. 15 ›› Issue (52): 9797-9802.doi: 10.3969/j.issn.1673-8225.2011.52.025

• 骨与关节图像与影像 bone and joint imaging • 上一篇    下一篇

结合各向异性扩散滤波Thin Plate先验正电子发射断层图像重建的算法

张  权，刘  祎   

  1. 中北大学电子测试技术国家重点实验室，山西省太原市   030051
  • 收稿日期:2011-06-07 修回日期:2011-07-21 出版日期:2011-12-24 发布日期:2011-12-24
  • 作者简介:张权★，男，1974年生，山西省大同市人，汉族，2005年上海大学毕业，硕士，讲师，主要从事图像处理、重建方面研究。 Zhangq2002@163.com
  • 基金资助:

    山西省自然科学基金重点项目(2009011020-2)；山西省高等学校科技开发项目资助(20081024)；中北大学2008年校青年科学基金。

Image reconstruction algorithm for positron emission tomography with Thin Plate prior combined with an anisotropic diffusion filter

Zhang Quan, Liu Yi   

  1. National Key Laboratory For Electronic Measurement Technology, North University of China, Taiyuan  030051, Shanxi Province, China
  • Received:2011-06-07 Revised:2011-07-21 Online:2011-12-24 Published:2011-12-24
  • About author:Zhang Quan★, Master, Lecturer, National Key Laboratory For Electronic Measurement Technology, North University of China, Taiyuan 030051, Shanxi Province, China Zhangq2002@163.com
  • Supported by:

    Key Project of Natural Science Foundation of Shanxi Province, No. 2009011020-2*; Program for Science and Technology Development in Shanxi Universities, No. 20081024*; Youth Science Foundation of North University of China in 2008*

PDF

373

摘要：

背景：在正电子发射断层成像中，MAP重建方法通过引入先验分布约束，可以明显提高重建图像的质量，但不合适的先验分布项可能会造成重建图像过度平滑或出现阶梯状边缘伪影。
目的：针对基于传统局部先验信息的MAP方法易于导致重建图像过平滑或产生阶梯状边缘伪影的问题，提出了一种结合各向异性扩散滤波的、基于Thin Plate先验的改进MAP重建算法。
方法：重建算法由两步组成：基于双向扩散系数的PDE各向异性扩散滤波和基于Thin Plate先验的MAP估计。重建图像通过这两步交替迭代得到。文中采用归一化均方根误差和信噪比定量评价重建图像质量。
结果与结论：结合了基于双向扩散系数的PDE各向异性扩散滤波，并将Thin Plate二次二阶先验模型引入到MAP重建算法中，所获得的重建结果图像在抑制噪声、边缘保持方面取得了良好的效果，SNR、RMSE以及视觉评价等方面均有较大程度的改善。

关键词: 正电子发射断层成像, 图像重建, Thin Plate先验, 各向异性扩, 散

Abstract:

BACKGROUND: In positron emission tomography imaging, maximun posterior (MAP) reconstruction can greatly improve the quality of reconstructed image by introducing prior distribution constraint. But a improper prior distribution may result in over-smoothess and stepladder edge of reconstructed image.
OBJECTIVE: To put forward an algorithm combines with anisotropic diffusion filter and MAP improved by Thin Plate prior according to over-smoothess and stepladder edge of reconstructed image by traditional MAP with local prior information.
METHODS: Reconstruction algorithm consists of anisotropic diffusion filter based on equation with forward-and-backward diffusion coefficient and MAP estimation based on Thin Plate prior. Reconstructed images were obtained by the alternate iteration of the above two steps. The quality of reconstructed images was evaluate by normalized rms error (RMSE) and signal-to-noise ratio (SNR).
RESULTS AND CONCLUSION: Reconstructed images obtained by MAP with second-order second Thin Plate prior model combined with anisotropic diffusion filter based on forward-and-backward diffusion coefficient partial differential equation were improved in restrain noise, edge-preserving, SNR, RMSE, visual evaluation and so on.

中图分类号: