Research On The Algorithm Of Artifact Reduction And Detail Preservation In Low Dose CT Images

X-ray Computed Tomography(CT)imaging technology has been widely used in clinical diagnosis since it was put forward.Compared with other radiological diagnosis methods,CT technology has the advantages of high resolution and low price,but excessive X-ray irradiation can lead to increased likelihood of the incidence of cancer,genetic disease and leukemia,and the dose of radiation will accumulate in patients.Therefore,it is important to control the radiation dose to obtain low-dose CT(LDCT)images.There are many ways to reduce the radiation dose,among which,lowering tube current is the simplest and most effective way,so it is also the most commonly used.Nevertheless,lowering tube current makes the projection data corrupted by quantum noise,making the corresponding reconstructed images badly damaged by mottle noise and streak artifacts,which interfere with the doctor’s diagnosis of the disease seriously.Therefore,it is of great clinical significance to improve the quality of CT images reconstructed under low-dose scanning condition.In order to improve the quality of LDCT images,this degree dissertation conducts in-depth research from following aspects: sparse representation theory,non-local means theory,morphological component analysis theory and partial differential equations method,etc.,and proposes three LDCT image domain post-processing algorithms.The main works are as follows:(1)To address the problem that the nonlocally centralized sparse representation(NCSR)algorithm suffers from residual streak artifacts and can’t preserve details information well when implemented in LDCT image denoising,and it has high computational complexity,in this degree dissertation,we propose an improved model,i.e.SNCSR model,based on the stationary principal component analysis(PCA)sub-dictionaries,NCSR and relative total variation(RTV).In the SNCSR model,in order to learn more accurate sub-dictionaries,the LDCT image is preprocessed by the improved total variation(ITV)model in which the weighted coefficient of the regularization term is constructed depending on a clipped and normalized local activity.In addition,the maximum eigenvalue of the gradient covariance matrix of the image patch is used to distinguish structure information from background region so that the restored image can be represented more sparsely.Moreover,unlike the NCSR model that needs to learn sub-dictionaries in each outer loop,the proposed model learns stationary sub-dictionaries only once before iteration starts,which shorten the computation time significantly.At last,the RTV algorithm is applied to reduce the residual artifacts in the recovered image more thoroughly.Compared with several other competitive denoising algorithms,the experiments performed on the simulated pelvis phantom,the actual thoracic phantom and the clinical abdominal data show that the proposed SNCSR model has lower computational complexity and can improve LDCT images quality more effectively.(2)The separation-based(SEPB)method is proposed for mottle noise and streak artifacts suppression in the LDCT images.In it,the LDCT image is decomposed into the structural image with residual mottle noise and the streak artifacts image with residual structural details by the image decomposition method RC.The structural image is filtered by the K-Singular Value Decomposition(K-SVD)algorithm to remove the residual mottle noise,and the structural details in the streak artifacts image are extracted by the morphological component analysis(MCA)theory.The extracted structural details are added to the filtered structural image to get the LDCT result image.Meanwhile,in the process of extracting the structural details,the streak artifacts dictionary learned from the streak artifacts image is corrected by the local intuitional fuzzy entropy,removing the structural atoms in it.The experiments are conducted on the modified Shepp_Logan phantom and the pelvis phantom to evaluate the effectiveness of the proposed SEPB method,and experimental results show that the SEPB method has better performance in subjective visual effect and objective indicators compared with several comparative denoising methods.(3)A novel total variation(NTV)model is proposed.A weighted coefficient of the regularization term of NTV model is constructed by standard deviation,gray-level probability and gradient magnitude to smooth LDCT images adaptively,since the standard deviation and the gray-level probability of detail region are higher than that of noisy background,and the gradient magnitude of edges is higher than that of noisy background.Besides,to preserve edges and details effectively,the fidelity term of the proposed NTV model is constructed by the block-matching 3D(BM3D)filter,because it performs well in details and edges preservation.The experiments are performed on the simulated Shepp_Logan phantom,the simulated pelvis phantom and the actual phantom.Compared with several other competitive methods,both subjective visual effect and objective evaluation criteria show that the proposed NTV model can improve LDCT images quality more effectively,such as noise and artifacts reduction,edges and details preservation.