计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (16): 4-6.

• 博士论坛 • 上一篇    下一篇

基于广义Gibbs先验的低剂量X-CT优质重建研究

马建华,黄 静,陈 阳,陈凌剑,陈武凡

  

  1. 南方医科大学 生物医学工程学院 医学信息研究所,广州 510515
  • 收稿日期:2008-01-08 修回日期:2008-03-14 出版日期:2008-06-01 发布日期:2008-06-01
  • 通讯作者: 马建华

Generalized Gibbs prior based high quality low-dose X-CT reconstruction

MA Jian-hua,HUANG Jing,CHEN Yang,CHEN Ling-jian,CHEN Wu-fan   

  1. Institute of Medical Information & Technology,School of Biomedical Engineering,Southern Medical University,Guangzhou 510515,China
  • Received:2008-01-08 Revised:2008-03-14 Online:2008-06-01 Published:2008-06-01
  • Contact: MA Jian-hua

摘要: 为获取低剂量条件下X-CT的优质重建,提出基于广义Gibbs先验的低剂量X-CT重建算法。新算法首先对投影数据进行统计建模,其后采用Bayesian最大后验估计方法,将投影数据中非局部的先验信息加诸于该数据的恢复中,达到抑制噪声的效果,最后仍采用经典的滤波反投影方法对恢复后的投影数据进行解释CT重建。文中将非局部先验称为广义Gibbs先验,其原因在于该先验具有传统Gibbs先验形式的同时,可以通过选择较大邻域和自适应的加权方式充分利用投影数据的全局信息进行数据恢复。通过与已有算法的对比实验,表明该文提出的基于广义Gibbs先验的低剂量X-CT重建算法在降低噪声效果和保持边缘方面具有较好的表现。

关键词: 低剂量X-CT, Bayesian估计, Gibbs先验, 广义Gibbs先验

Abstract: In order to obtain the high quality reconstruction images for low-dose X-CT,we proposed a new generalized Gibbs prior based low-dose X-CT reconstruction method.First,X-CT projection data was modeled as a statistics process.Then based on Bayesian maximum posteriori estimation method,we designed a novel Gibbs prior named as generalized Gibbs prior,which exploited nonlocal information of the data to suppress noise.Last,we used the filtered back-projection method to finish the final CT reconstruction.The reason for the name of generalized Gibbs prior is that it has been shown to suppress noise effectively while capturing sharp edges without oscillations through the selection of larger neighborhood and adaptive weight form with global information of an image.Comparisons between the new method and other models clearly demonstrate that the proposed generalized Gibbs prior based low-dose X-CT reconstruction method performs better in lowering the noise and preserving the edge and detail in the image.

Key words: low-dose X-CT, Bayesian estimation, Gibbs prior, generalized Gibbs prior