Performance of sparse-view CT reconstruction with multi-directional gradient operators

Abstract

To further reduce the noise and artifacts in the reconstructed image of sparse-view CT, we have modified the traditional total variation (TV) methods, which only calculate the gradient variations in x and y directions, and have proposed 8- and 26-directional (the multi-directional) gradient operators for TV calculation to improve the quality of reconstructed images. Different from traditional TV methods, the proposed 8- and 26-directional gradient operators additionally consider the diagonal directions in TV calculation. The proposed method preserves more information from original tomographic data in the step of gradient transform to obtain better reconstruction image qualities. Our algorithms were tested using two-dimensional Shepp–Logan phantom and three-dimensional clinical CT images. Results were evaluated using the root-mean-square error (RMSE), peak signal-to-noise ratio (PSNR), and universal quality index (UQI). All the experiment results show that the sparse-view CT images reconstructed using the proposed 8- and 26-directional gradient operators are superior to those reconstructed by traditional TV methods. Qualitative and quantitative analyses indicate that the more number of directions that the gradient operator has, the better images can be reconstructed. The 8- and 26-directional gradient operators we proposed have better capability to reduce noise and artifacts than traditional TV methods, and they are applicable to be applied to and combined with existing CT reconstruction algorithms derived from CS theory to produce better image quality in sparse-view reconstruction.

Multidirectional Anisotropic Total Variation and Its Application in the Tomography of the Surrounding Rock of Coal Mining Faces

“…Previous studies [3,4] reported that the ITV uses the multidirectional gradient information of pixels and showed that multidirectional ITV could effectively improve imaging accuracy. Inspired by their research, we propose a novel ATV that uses the 8-direction gradient information of a pixel, that is, MATV.…”

“…X-ray computed tomography (CT) has been widely used in medical diagnosis [1][2][3][4] since it was invented in 1973. X-ray CT scans automatically explore the internal structure of the human body.…”

“…Sidky and Pan [2] introduced TV into medical X-ray CT for the first time. According to the existing research [3,4], the accuracy of this kind of algorithm is higher than the accuracy of a direct iterative method.…”

“…Researchers around the globe have made numerous attempts to suppress staircase artifacts and improve imaging accuracy. ese improvements can be divided into the following categories: (1) generalize the difference of operation in TV to a higher-order difference [18][19][20] or fractional-order difference [21] such as high-order TV [18][19][20][21], total generalized variation [22], and fractional-order TV [23,24]; (2) extend the L1 norm in TV to the Lp quasinorm (0<p < 1) that can obtain sparser solution than the L1 norm, such as TpV [11,17,25,27]; (3) generalize the gradient in TV to the overlapping combination gradient information, such as overlapping group sparsity total variation [28][29][30][31][32]; (4) use Tikhonov regularization and TV regularization to solve the objective function in an alternative way, such as modified TV [33][34][35]; and (5) generalize the two-directional gradient to multidirectional gradient in TV, such as multidirectional TV [3,4,36]. TV can be divided into isotropic total variation (ITV) and anisotropic total variation (ATV) [10,37].…”

Provably Convergent Learned Inexact Descent Algorithm for Low-Dose CT Reconstruction

“…As high dosage X-rays can be harmful to human body [3,4,5], substantial efforts have been devoted to image reconstruction using low-dose CT measurements [6,7,8]. There are two main strategies for dose reduction in CT scans: one is to reduce the number of views, and the other is to reduce the exposure time and the current of X-ray tube [9], both of which will introduce various degrees of noise and artifacts and then compromise the subsequent diagnosis. Here we focus on the second type however our method is not specific to a particular scanning mode.…”