Research On CT Reconstruction Algorithm Applied To Oral Cavity

Computed tomography(CT)is the cornerstone of modern radiography.As the basic principle of CT,image reconstruction aims to reconstruct objects from the measured projection data.The purpose of various image reconstruction algorithms is to produce high quality reconstructed images as soon as possible.Generally speaking,image reconstruction algorithm can be divided into analytic algorithm and iterative algorithm.Analytic algorithm is considered to be faster than iterative algorithm,which is gradually used in most advanced commercial CT scanners.The main contents are as follows:First,this paper introduces the mathematical relationship between filtered back projection(FBP)and differential back projection(DBP).Using the mathematical relationship between two typical analytical reconstruction algorithms,six analytical reconstruction formulas are given.The main evaluation methods of reconstructed image quality are described,which paves the way for the follow-up research of this subject.Secondly,this paper proposes a method to improve the effect of suppressing CT metal artifacts by reducing the influence of interpolation error,which reduces metal artifacts and improves the quality of reconstructed image.In the filter back projection algorithm,a non-local ramp filter is used to filter the interpolation projection data before back projection,so that the interpolation error can be transmitted to the whole projection data.The metal artifact suppression method based on the differential back projection reconstruction algorithm can filter after the back projection along the preferred Hilbert filter direction.This attribute helps to reduce the propagation of interpolation error and improve the quality of reconstructed image.At the same time,numerical simulation experiments are used to verify the performance of the algorithm.Thirdly,in order to reduce the X-ray radiation and scan time,this paper studies the CT iterative reconstruction algorithm using reference image.First,the Lagrangian dual method is used to transfer the constrained minimization problem to the corresponding unconstrained maximization dual problem.Then,the distance function composed of the norm of the reconstructed image and the reference image is used as the cost function of image reconstruction.Finally,the gradient rise method is used to iterate many times to optimize the cost function.At the same time,four reference image selection schemes are proposed.The experimental results show that the quality of the reconstructed image can be improved by choosing the right reference image with less iterations.Among them,the FBP reconstruction image of reference image is easy to obtain and does not need to know the expected structure of the scanning object;the image quality of the reference image is significantly improved,with the least distortion,which is the closest to the original image,basically without stripe artifacts,and more importantly,the details of the original image are retained.