Research On CT Image Reconstruction Algorithms Using Undersampled Projection Data

Computed tomography(CT),by virtue of its superiority,is widely used in the medical field.However,excessive X-ray radiation dose to the patients may potentially increase the risk of cancer.Under the premise of not changing the existing hardware equipment,sparse-view and limited-angle scanning can effectively reduce the CT radiation dose.Due to the insufficient of projection data,the quality of reconstructed image degrades seriously.Therefore,how to reduce radiation dose while ensuring the quality of reconstructed images is a hot research topic in the current medical imaging domain.This paper mainly focuses on the problem of sparse views and limited angle image reconstruction in low dose CT imaging.The research background and the development of low-dose CT image reconstruction were introduced firstly in this paper.Secondly,the theoretical foundation of CT imaging and classical reconstruction algorithms are introduced.Meanwhile,three image reconstruction methods are proposed and implemented as following:Firstly,an adaptive non-local means(NLM)based reconstruction method for sparse-view CT is proposed to overcome the problem of over_smoothness at image edges in traditional ART-NLM algorithm,named as ART-ANLM.In ANLM,a novel similarity window is defined and a similarity measure based on rotation transformation is designed so that the similarity between two pixels can be measured accurately to avoid blurring of image edge.Moreover,a new filter parameter in ANLM,which controls the decay of the weights would change adaptively with the iteration number increases,and vary with the pixel gradient.This is another way to avoid over-smoothed image.The proposed method is validated on simulated and real projection data.The reconstructed results demonstrated that the ART-ANLM method could remove artifacts or noises more effectively and preserve image edges better.Compared to the conventional ART-NLM method,the SNR and MAE from ART-ANLM increases 38%and decreases 76%,respectively.Secondly,an adaptive NLM reconstruction algorithm based on rotation invariance(ART-RIANLM)is presented to solve the problem of over-smoothness at image edges in traditional ART-NLM algorithm for sparse-view CT reconstruction.In RIANLM,a similarity measure based on rotational invariance is proposed to calculate the distance between two similarity windows.In this way,any patch with similar structure but different orientation to the reference patch would win an appropriate weight to avoid over-smoothness.Moreover,the filter parameter h in RIANLM which controls the decay of the weights is adaptive to avoid over-smoothed image.The method is validated on simulated and real projection data.The experimental results indicate that the ART-RIANLM algorithm can achieve a better trade-off between removing artifacts or noises and preserving edges.Compared to the conventional ART-NLM method,the SNR and MAE from ART-RIANLM increases 51%and decreases 74%,respectively.Thirdly,a limited-angle CT image reconstruction algorithm based on local and non-local regularization is proposed to deal with the problem of geometric distortion artifacts in limited-angle CT image reconstruction.By modifying the traditional NLM algorithm,the image information of the non-artifact region is used to recover the pixel value of the artifact region,and then the total variation minimization method is used to eliminate the false structure information brought by the non-artifact region.Simulated and real projection data are used to validate the feasibility of this algorithm.The experiment results show that the proposed method can greatly reduce the geometric distortion artifacts of reconstructed image,and improve the image quality significantly.Compared to ART-NLM method,the SNR and MAE from ART-mNLM/TV increases 57%and decreases 52%,respectively.In this paper,three improved reconstruction algorithms are proposed for sparse and limited-angle CT image reconstruction,which significantly improve the reconstructed image quality.Some preliminary achievements in image quality have been obtained in this paper,but it still needs to be further explored and perfected for practical clinical application.