An iterative method for simultaneous reduction on beam-hardening and scatter artifacts in x-ray CT. (1st April 2023)

• Record Type: Journal Article
• Authors: Xia, Dimeng; Zhao, Shusen; Zhang, Huitao; Zhu, Yining; Zhao, Xing; Zhang, Peng
• Is Part Of: Physica scripta. Volume 98:Number 4(2023)
• Publication Date: 2023-04-01
• Subjects: beam-hardening artifact reduction -- scatter artifact reduction -- deconvolution -- iterative method -- x-ray CT; Physics -- Periodicals; 530.05
• DOI: 10.1088/1402-4896/acc61b
• Languages: English
• ISSNs: 0031-8949

Abstract

Abstract: Beam-hardening and scatter are two significant factors leading to contrast reduction and gray value inaccuracy in CT images. The cupping artifacts and obscure boundaries in reconstructed images are also caused mainly by both beam-hardening and scattering. It is difficult to get high-quality CT images with only one of them to make correction. In this paper, we proposed an x-ray CT polychromatic attenuation model with scatter effect, and an iterative method for simultaneous reduction on beam-hardening and scatter artifacts. In this model, the measurements of the detector comprise two parts: an attenuation term and a scatter term. The former is defined by an exponential rational fraction to fit the traditional attenuation process, and the latter is defined by a convolutional scatter intensity. The coefficients of the rational fraction in the attenuation term and the scatter term kernel are all calculated from a calibration phantom which is scanned to get corresponding equations. Based on the polychromatic attenuation model, we proposed an iterative artifacts reduction method combining deconvolution technique with linearized back-projection (iDLB method). This method makes the nonlinear polychromatic attenuation model become easily solvable by linearizing transformation, which simplifies the residuals allocation process. Experiments of both numerical simulation and practical data show the iDLB method has the ability to reduce beam-hardening and scatter artifacts simultaneously, improve the contrast of CT images, and it is highly parallelized for lower computational cost.