The Research Based On Statistical Iterative Reconstruction Algorithm For Low-dose Computed Tomography

With the rapid development of CT, in the study of algorithms principle either, or in termsof actual clinical applications, computed tomography have made remarkable achievements tobecome one of the most accurate and the most potential method. Although CT examination iswidely recognized in clinical diagnosis, but when the X-ray radiates to living organisms, dueto radiation-induced lesions occur in the body, it will cause damage in varying degrees.X-ray radiation issue of patients related to CT scans has aroused widespread concern andattention. Therefore, more and more scholars began to study the low-dose CT technology.Low-dose CT image reconstruction, in essence, is a problem of noise suppression. The mainmethods of low-dose CT reconstruction are noise reduction processing of projection data anditerative reconstruction algorithm based on statistics. Because filtered back-projectionreconstruction algorithm is sensitive to noise, reduce radiation dose will lead to thedegradation of reconstruction image quality, and the increase of noise. Compared to filteredback projection reconstruction algorithm, in the conditions of more noisy, iterativereconstruction technique based on statistical can obtain high SNR reconstruction results,dramatically reducing the radiation dose in CT scan. This paper describes the CTreconstruction on the basis of principles and classical algorithms, mainly for low-dose CTnoise model based on low-dose CT and statistical iterative reconstruction algorithm has beenstudied mainly as follows:1. First describe the physical principles and basic mathematics about CT reconstructionalgorithms; then focus on the CT reconstruction of several types of classical algorithms,including filtered back projection reconstruction algorithm, algebraic reconstructionalgorithm. Filtered back projection reconstruction algorithm has the advantage of fast reconstruction, to meet the real-time, but because it is more sensitive of the noise and artifacts,there are some restrictions to reduce radiation dose. Algebraic reconstruction algorithm is notwidely used, due to it did not fully consider the statistical properties of the noise model inlow-dose CT and its convergence is for a long time, accuracy and noise suppression are notgood. Finally, according to the characteristics of X-ray photons, the analysis of noise modelin low-dose CT reconstruction process provids the foundation for the study of statisticaliterative reconstruction algorithm later.2. For degradation problems of low-dose CT reconstructed images due to excessivequantum noise, based on statistical reconstruction algorithm, this paper proposes a newpunishment weighted least squares image reconstruction algorithm applied to the imagedomain, to solve the noise problems in low-dose CT reconstruction process. First, through theanalysis of the mechanism of partial differential equations and PM diffusion model, for thedrawback of traditional diffusion model considering image gradient only, an new diffusionoperator has been proposed, then the traditional cost function in PWLS algorithm has beenimproved by the introduction of various anisotropic diffusion operator, such that thealgorithm can be based on the inherent characteristics of the image, adaptively adjust itsdegree of smoothing, so that the reconstruction algorithm can effectively suppress imageartifacts strip, but also protect the edges and details of the reconstructed image.3. In the statistical iterative reconstruction algorithm, maximum likelihood expectationmaximization (MLEM) algorithm due to take full account of the imaging process of thephysical model, and the statistical properties of observed data, also a very important feature isthat it is non-negative, in practice it has been widely used. In this paper, by the combining ofmathematical theory of fuzzy and bilateral filtering thinking, a new diffusion model forregularization MLEM reconstruction algorithm have been proposed to achieve noisesuppression effect. Experimental results show that, at a more noisy conditions of low dose CT,the algorithm is still able to reconstruct the optimum quality images.