基于投影数据全广义变分最小化的低剂量CT重建
发布时间：2018-09-08 09:10
【摘要】：目的提出基于投影数据全广义变分最小化的低剂量CT重建方法。方法首先,通过非线性Anscombe变换将满足Poisson分布的投影数据转化为近似Gaussian分布,然后基于全广义变分正则化模型对变换后的Gaussian型数据进行噪声去除。最后,对去噪的结果进行Anscombe逆变换后实现传统的滤波反投影(FBP)CT重建。结果数值体膜实验结果表明本文提出的方法可以大大地改进重建图像的质量。FBP方法重建的Clock和Shepp-Logan体膜图像的信噪比分别为17.752 dB和19.379 dB,本文方法重建的图像的信噪比提高到24.0352 dB和23.4181 dB。FBP方法重建方法重建的Clock和Shepp-Logan体膜图像的均方误差分别为0.86%和0.58%,本文方法重建的图像的均方误差降低到到0.2%和0.23%。结论本文方法可以在投影数据不满足分段常数假设的前提下去除噪声和条形伪影,从而提高低剂量CT图像重建质量。
[Abstract]:Aim to propose a low dose CT reconstruction method based on full generalized variational minimization of projection data. Methods firstly, the projection data satisfying Poisson distribution is transformed into approximate Gaussian distribution by nonlinear Anscombe transform, and then the transformed Gaussian type data is denoised based on the fully generalized variational regularization model. Finally, the Anscombe inverse transform is applied to the denoising result to realize the traditional filtering backprojection (FBP) CT reconstruction. Results the results of numerical volumetric membrane experiments show that the proposed method can greatly improve the quality of reconstructed images. The SNR of Clock and Shepp-Logan films reconstructed by the proposed method is 17.752 dB and 19.379 dB, respectively. The signal-to-noise ratio of the images reconstructed by this method is 17.752 dB and 19.379 dB, respectively. The mean square error of Clock and Shepp-Logan images reconstructed by the method of 24.0352 dB and 23.4181 dB.FBP is 0.86% and 0.58%, respectively. The mean square error of the reconstructed image is reduced to 0.2% and 0.23% respectively. Conclusion the proposed method can remove noise and strip artifacts without satisfying the piecewise constant assumption of projection data, thus improving the reconstruction quality of low-dose CT images.
【作者单位】： 赣南师范大学数学与计算机科学学院;
【基金】：国家自然科学基金(11701097,11661007,11661008) 江西省自然科学基金青年基金(2016BAB212055) 江西省教育厅科学技术研究项目(GJJ150994)~~
【分类号】：R814.42;TP391.41
本文编号：2230084
[Abstract]:Aim to propose a low dose CT reconstruction method based on full generalized variational minimization of projection data. Methods firstly, the projection data satisfying Poisson distribution is transformed into approximate Gaussian distribution by nonlinear Anscombe transform, and then the transformed Gaussian type data is denoised based on the fully generalized variational regularization model. Finally, the Anscombe inverse transform is applied to the denoising result to realize the traditional filtering backprojection (FBP) CT reconstruction. Results the results of numerical volumetric membrane experiments show that the proposed method can greatly improve the quality of reconstructed images. The SNR of Clock and Shepp-Logan films reconstructed by the proposed method is 17.752 dB and 19.379 dB, respectively. The signal-to-noise ratio of the images reconstructed by this method is 17.752 dB and 19.379 dB, respectively. The mean square error of Clock and Shepp-Logan images reconstructed by the method of 24.0352 dB and 23.4181 dB.FBP is 0.86% and 0.58%, respectively. The mean square error of the reconstructed image is reduced to 0.2% and 0.23% respectively. Conclusion the proposed method can remove noise and strip artifacts without satisfying the piecewise constant assumption of projection data, thus improving the reconstruction quality of low-dose CT images.
【作者单位】： 赣南师范大学数学与计算机科学学院;
【基金】：国家自然科学基金(11701097,11661007,11661008) 江西省自然科学基金青年基金(2016BAB212055) 江西省教育厅科学技术研究项目(GJJ150994)~~
【分类号】：R814.42;TP391.41
【相似文献】
相关期刊论文 前2条
1 林友俊;;核磁共振的应用与现状[J];铁道医学;1990年01期
2 徐昊,庄天戈,柴新禹;基于点探测器和柱面源双位置扫描的直接体积CT的研究[J];电子学报;2002年10期
相关博士学位论文 前4条
1 上官宏;低剂量X线CT统计迭代重建方法研究[D];中北大学;2016年
2 齐宏亮;有限投影数据CT图像迭代重建技术研究[D];南方医科大学;2017年
3 毛宝林;低剂量CT图像重建与伪影校正算法研究[D];东北大学;2015年
4 张华;基于系统建模的低剂量CT重建研究[D];南方医科大学;2014年
，本文编号：2230084
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2230084.html
