Low-dose CT Imaging Reconstruction Via Total Generalized Variation Regularization

Nowadays,Computed Tomography(CT)technology has been widely applied in medical diagnosis and treatment,people are increasingly concerned about the effect of X-ray radiation dose on human health.There is always a trade-off between X-ray radiation dose and the CT images quality.The higher the radiation dose,the better the image quality.However,excessive radiation exposure during clinical examinations has been reported to be linked to increased lifetime risk of cancers in patients.Therefore,low-dose CT is under the spotlight in recent years.Reducing tube current or tube voltage in CT scan is the simplest and effective method to achieve low-dose imaging.However,sinogram data will be contaminated by quantum noise and electronic noise.As a consequence,the quality of reconstructed CT image from the conventional filtered back-projection(FBP)will be severely degraded.Therefore,it is highly desirable to minimize of X-ray radiation dose while ensuring image quality in the field of CT imaging.There are many low-dose CT reconstruction methods proposed,which can be roughly disjoined three categories as follows: one is the image post-processing method,which directly post-processes the generated image without considering the source of image noise.The other is the low-dose CT reconstruction method based on the image domain.The third is the low-dose CT reconstruction method based on the projection domain.The low-dose CT reconstruction method based on projection domain not only has fast calculation speed,but also can effectively reduce noise and strip artifacts.Therefore,this paper will be combined the optimal algorithm to improve the CT images quality by adding regularization prior information.Next,this paper main research work will be introduced as follows:1.Firstly,this paper describes the development background and research progress of CT imaging,then explains three common image reconstruction algorithms,and finally displays the noise model used in the paper and uses the optimization algorithm to solve it.2.In view of the serious degradation of image quality in low-dose CT reconstruction,we construct a statistical model according to the statistical characteristics of CT projection data and introduce regularization prior information for low-dose CT image reconstruction.However,total variation(TV)regularization can effectively reduce noise while maintaining the spatial resolution of the image under the premise of satisfying the piecewise constant of image,but it is easy to produce the noticeable ladder effect and patchy artifacts in the reconstructed CT images.Therefore,we introduce TGV regularization as prior information into the recovery process of projection data,and propose a total generalized variation constrained weighted least squares(TGV-WLS)for low-dose CT reconstruction method,which can achieve noticeable gains in terms of noise-induced artifacts suppression and edge detail preservation without satisfying the assumption of piecewise constant of image.3.TGV regularization not only utilizes higher order derivatives of the objective image,which can yield visually pleasant images without satisfying the assumption of piecewise constant,but also TGV regularization can apply to the radon domain.Based on these advantages,we propose a statistical iterative reconstruction for low-dose CT using spatial-radon domain total generalized variation constrained penalized weighted least-squares(PWLS-SRDTGV)to better enhance the image quality of low-dose CT.First,an iterative image reconstruction algorithm based on a penalized weighted least-squares(PWLS)principle was used to reduce the noise,and then the reconstructed image and its corresponding higher resolution projection by penalizing of their TGV regularization coefficients simultaneously.Subsequently,an effective split Bregman algorithm is applied to solve the associated objective function in the PWLS-SRDTGV reconstruction.Qualitative and quantitative studies were conducted to evaluate the presented PWLS-SRDTGV method using a digital XCAT phantom and physical phantom.The experimental results show that PWLS-SRDTGV method can achieve noticeable gains in terms of noise-induced artifact suppression and edge detail preservation compared with other competing methods.