Low Dose CT Image Reconstruction Based On Image Quality Assessment

As X-ray Computed Tomography(CT)plays an important role in modern medical science,the problem of its radiation becomes a focal spot of public.It is well known that X-ray is harmful to human body.As a result,low does CT gets more attention in modern society concerning human tolerance to radiation.There are two ways to achieve low dose CT:decreasing current intensity(low current intensity)and reducing the number of projections(sparse views).Under the former condition,more noise would be detected in projections received by detectors when current of CT X-ray source drops,which leading to noisy reconstructed images.Under the later condition,classical reconstruction method would generate images with streak artifacts because of under sampling.Both of these two issues would affect the quality of CT image reconstruction.It is a general way to deal with low current intensity issue by suppressing noise in the noisy projections.However,these noise reduction methods are often complex and with low computational efficiency.Thus,it cannot be applied in practical environment which requires high calculation speed.Moreover,this kind of methods cannot eliminate streak artifacts because of sparse views.To solve this problem,iterative reconstruction(IR)based on compress sensing theory seems the optimal solution.Unfortunately,IR is time-consuming and also cannot handle with the problem of huge quality loss derived from noise in projections.In this paper,for solving the problems in traditional CT reconstructions,a parallel projection noise reduction method and an iterative reconstruction algorithm framework are put forward by investigating the relationship among CT image quality,X-ray dose and reconstruction method based on image quality.Three points are shown as follows:(1)Based on topographic independent component analysis(TICA),this paper advances a no-reference image quality assessment(IQA)method.After extracting features related to the inherit quality of CT image and comparing the bias of characteristic distribution,TICA is successful to evaluate the quality of CT image.During experiment,two databases are built up for studying CT image quality.By using the IQA algorithm mentioned,some conclusions are drawn on the relationship among CT image quality,X-ray dose and reconstruction algorithm.(2)Based on relative IQ A,this paper designs a parameter optimization system for a parallel projection noise reduction method.By modeling the properties of CT noise in projections,a parallel iterative operator formula is created which improves the computational efficiency.Meanwhile,a parameters optimization system based on IQA is designed for better performance.In the experiments of this paper,the performance of the proposed algorithm upgrades 16%at least comparing to other classical algorithms.(3)By considering results of image quality to converge,this paper describes a framework of an iterative reconstruction algorithm which based on TV(Total Variation)minimum.During the iteration of the algorithm,the convergence judgment is the image quality of every result after iteration,which accelerating the convergence rate of IR algorithm.The experimental results show that the performance and efficiency of the proposed algorithm are both improved.Also,the adaptability and robustness of this framework to low dose CT are progressed obviously.