基于梯度保真项的低剂量CT统计迭代重建算法

Statistical Iterative Reconstruction Algorithm Based on Gradient Fidelity for Low-dose CT

针对低剂量计算机断层扫描(Computed Tomography,CT)重建图像时容易出现明显条形伪影这一现象,提出一种基于梯度保真项的低剂量CT统计迭代重建算法。该算法克服了原始全变分(Total Variation,TV)模型在抑制条形伪影和噪声的同时引入阶梯效应的缺点,首先把梯度保真约束项和能够区分图像平滑区和细节区的边缘指示函数应用到TV模型中得到基于梯度保真项的自适应全变分模型,然后再把新模型与惩罚加权最小二乘(Penalized Weighted Least Square,PWLS)重建算法相结合,使用交替方向迭代法得到最终的图像。采用Shepp-Logan模型来验证算法的有效性,实验结果表明,该算法不仅可以有效地去除条形伪影,还可以较好地保护图像的边缘和细节信息。

For the phenomenon of obvious streak artifacts when the low-dose computed tomography（ CT） reconstructs images,a low-dose CT statistical iterative reconstruction method based on gradient fidelity term is presented. This method overcomes the shortage that as the traditional total variation（ TV） suppress streak artifacts and noises; it has to bring staircase effect at the same time. Firstly,the paper applies the gradient fidelity constraint entry and the edge indicator function which can distinguish image smoothing area and detail area to the TV model,and then it gets adaptive total variation model based on gradient fidelity term. After that,it combines the new model with the penalized weighted least square（ PWLS） and gets the final image by using alternating direction iteration method.The Shepp-Logan model is used to verify the effectiveness of this algorithm. The experimental results show that the proposed algorithm can not only reduce the streak artifacts effectively,but also preserve the image edges and details information very well.