Image Reconstruction Methods For Low-dose Cone-beam Computed Tomography

Cone Beam CT is one type of Computed Tomography device,the data of which is acquired by rotating X-ray generator around object for digital radiography.Reconstructing the intersection data by computational methods generates three-dimensional images.The principle of data acquisition in cone-beam CT is different compared with conventional CT;however,they both share similar reconstruction algorithms.The main difference between cone beam CT and conventional CT is the projection data dimension,in in which the former one is three dimensional images reconstructed by two dimensional data,and the latter is two dimensional image reconstructed by one dimensional data.The rebuild three-dimensional images in conventional CT are image stacks build by 2D data,which leads to heavy mental artifacts.Moreover,on the structural basis,cone beam CT utilizes cone beam instead of fan beam and the corresponding flat panel detector to replace conventional detector.Evidentially,cone beam x-ray scan can increase x-ray utilization by only one round of rotation to acquire all data,and flat panel detector could speed up data acquisition process.Another advantage for CBCT is its high spatial resolution.X-ray could increase the risk of many types of cancer,especiallypediatrics and women are sensitive to x-ray irradiation.Cone beam CT is used for multiple fraction of scan for orthodontics,radiation oncology and image guided surgery.Thus,strategies for lower the CBCT radiation dose such as lower the current or sparse scan protocol are essential for reducing x-ray risks.However,low dose often accompaniesnoise and artifacts in images.With the advance of computer technologies and the introduction of new mathematical models in image reconstruction,denoising and restoration,new methods are applied for low dose imaging in cone beam CT.In this article we analyze CBCT imaging principles and compressed sensingtheory in lowdose reconstruction,and solve the low dose CBCT image reconstruction model using convex optimization.This article includes three following aspects focusing on iterative regularization term and minimization of reconstruction function of low-dose CBCT image reconstructions:(1)We introduce a Block Matching and 3D filtering(BM3D)frame to CBCT image reconstruction based on block matching principle.Using the priori image and pre-reconstruction image for image registration to construct block-matching-dictionary.Contrary to the original BM3D methods,the frame-based version use matrix based computation as regularization term.The new term inherits the original BM3D filtering advantages and could better protect image edges and better suppress noise in smooth area compared with total variation and tight frame.Experiments reveals that the new methods could increase image contrast to noise ratio(CNR)and spatial resolution with edge protection and noise reduction.(2)An improved cone beam reconstruction method based on split-Bregman.The new method is based on Bregman distance and Bregman iteration,using splitting techniques to separate L1 regularization term and L2 data fidelity term in the minimization function.Experiments show that the adopted split-Bregman method could acquire better image quality and computational time compared with conventional projection-onto-convex-set(POCS)method with regularization,and could be wildly adopted.(3)An improved fused analytical and iterative reconstruction method for low dose CBCT reconstruction.The new method could combine the advantages of both analytical and iterative reconstruction,and introduce filtered back projection(FBP)to the fidelity term as a pre-conditioner.Experiments show that the fused analytical and iterative reconstruction algorithm could enhance image contrast and spatial resolution and acquire the state-of-the-art image quality.