| Cone-beam Computed Tomography(CBCT)is widely used in various specialized fields,such as image-guided radiation therapy,oral imaging,extremity imaging,and intraoperative diagnosis,due to its advantages of low radiation dose,high spatial resolution,and flexible scanning.However,the physical imaging characteristics of CBCT lead to a serious problem of photon scattering during imaging.The scatter radiation cause a decrease in soft tissue contrast,inaccurate CT values,and introduces cupping and streaking artifacts in reconstructed image.Therefore,solving the scatter problem is crucial to improve the quality of soft tissue imaging and promote the clinical application of CBCT.At present,CBCT scatter correction methods can be divided into three main categories:hardware-based,software-based,and hybrid correction.Monte Carlo(MC)method is one of the software-based correction methods and is the gold standard for particle transport.It can accurately estimate the scatter distribution to realize scatter correction.However,current MC-based CBCT scatter correction mainly faces the following problems:(1)low simulation efficiency.Conventional MC spends a lot of time on transporting more than 80%of photons that are unrelated to the calculation of scatter distribution.(2)High time cost.Due to the inaccurate CT values of CBCT images,iterative MC simulation is required to obtain accurate scatter distribution for correction,which further increases the time burden of this method.To achieve fast and accurate CBCT scatter correction,a correlated sampling-based Monte Carlo simulation method is proposed in this thesis,which mainly focuses on its implementation,technical process and application scope.First,a MC scatter estimation method based on correlated sampling variance reduction technique is proposed and an iterative correction framework is built.Due to the strong correlation of MC sampling system during iterative simulation,reusing the effective sampling space in the reference sample can reduce the track of task-unrelated photons.Correlation samples are forced to use the same set of random number sequences as the reference sample.Then a global correction factor is used to correct the result due to CT value perturbation in MC simulation system,which can improve the efficiency of the MC scatter simulation and achieve fast scatter correction.The results of simulation and measurement experiments show that MC simulation with correlated sampling can achieve efficient and unbiased scatter estimation.A significant improvement in CT value accuracy and image contrast are shown in the corrected CBCT images.Compared with the conventional MC,the efficiency gain brought by correlated sampling is between 23-43 times.The proposed method takes less than 25 seconds for the whole iterative scatter correction process.Second,further exploration is conducted on the applicable conditions of correlated sampling-based scatter correction method when geometric deformation occurs in the MC simulation samples.The concept of deformation variation coefficient is proposed to measure the correlation of the MC sampling system under image deformation.The deformation field of the correlation sample is obtained using the fast deformation registration Demons algorithm,and the deformation coefficient of variation is obtained by weighting the deformation distance in three-dimensional space using a linear attenuation coefficient.Then,the applicable conditions of the correlated sampling are determined by combining the results of scatter correction.The experimental results show that when the coefficient of variation deformation is within 0.3,the mean absolute percentage error value of the scatter distribution obtained by correlated sampling is within 3%.The proposed method can effectively remove the scatter artifacts from CBCT after correction,which has certain reference significance for expanding the application of correlated sampling. |
PDF Full Text Request