Low-Dose X-ray CT Reconstruction Algorithm Using Shearlet Sparse Regularization

Abstract

Computed tomography (CT) scans produce ionizing radiation in the body, and high-dose CT scans may increase the risk of cancer. Therefore, reducing the CT radiation dose is particularly important in clinical diagnosis, which is achieved mainly by reducing projection views and tube current. However, the projection data are incomplete in the case of sparse views, which may cause stripe artifacts in the image reconstructed by the filtered back projection (FBP) algorithm, thereby losing the details of the image. Low current intensity also increases the noise of the projection data, degrading the quality of the reconstructed image. This study aimed to use the alternating direction method of multipliers (ADMM) to address the shearlet-based sparse regularization problem, which is subsequently referred to as ADMM-shearlet method. The low-dose projection data were simulated by adding Gaussian noise with zero mean to high-dose projection data. Then FBP, simultaneous algebraic reconstruction technique, total variation, and ADMM-shearlet methods were used to reconstruct images. Normalized mean square error, peak signal-to-noise ratio, and universal quality index were used to evaluate the performance of different reconstruction algorithms. Compared with the traditional reconstruction algorithms, the ADMM-shearlet algorithm performed well in suppressing the noise due to the low dose while maintaining the image details.