Low-Dose X-ray Computed Tomography Reconstruction Using Curvelet Sparse Regularization

This study aimed to propose a mathematical problem based on the curvelet framework and used the alternating direction method of multipliers (ADMM) to solve this problem so as to achieve the purpose of image reconstruction. The method was called ADMM-Curvelet in this study. The ADMM-Curvelet algorithm was used for reconstruction. The required low-dose (i.e., 17, 40, 60, and 80 mAs) projection data were obtained by simulation. The ADMM-Curvelet algorithm was used to reconstruct the projection data from different mAs and projection views. The peak signal-to-noise ratio (PSNR) and normalized mean square error (NMSE) were used to evaluate the ability of the reconstruction algorithm to suppress noise. Moreover, the universal image quality index (UQI) was used to evaluate the quality of the image, namely, the similarity between the reconstruction and the original images. The maximum PSNR and the smallest NMSE could be obtained using the ADMM-Curvelet algorithm. The UQI value obtained by the ADMM-Curvelet algorithm was close to 1. The evaluation of this method clearly showed that the curvelet sparse regularization was more advantageous than the total variation and traditional reconstruction algorithm (i.e., algebraic reconstruction technique and filtered back projection) in suppressing noise and improving image similarity. Iterative reconstruction using curvelet sparse regularization could significantly improve the quality of the reconstructed image.