The Study Of CT Statistical Reconstruction Algorithm Algorithm Based On Maximum Likelihood And Likelihood And Penalized Likelihood Estimates

When the projection data has serious noise in the process of acquisition system and is incomplete, analytic reconstruction algorithms get images with artifact. The statistical reconstruction algorithm with a accurately physical model isn't sensitive to noise and easy to add constraints etc.Therefore, the reconstructed image quality is superior to conventional FBP methods. For the CT Statistics reconstruction, this content is as follow:The maximum likelihood estimate for the statistical reconstruction algorithms is studied and mainly includes expectation maximum algorithm (ML-EM),separable paraboloidal surrogate(ML-SPS) algorithm and the both algorithms with ordered subset(OS-ML-EM and OS-ML-SPS). In simulation experiments, it shows that the initial convergence rate of ML-SPS is faster than the ML-EM, however, the both algorithms is slower than OS-ML-EM and OS-ML-SPS algorithms. At last, OS-ML-SPS and OS-ML-EM algorithms are used to reconstruct actual CT projection data and get better image.Penalized maximum likelihood(PL) reconstruction for the statistical algorithms is based on the penalized likelihood estimates and adds a penalty term to suppress noise, which mainly includes OSL-EM algorithm and PL-SPS algorithm. It is focused on the penalty function which is based on Gibbs distribution and analyzes the need of potential functions to meet the conditions, the influence on image for the selection of potential functions and the advantages and disadvantages of the quadratic function and the Huber function. At last, simulation experiments give the error analysis of reconstructed image.This method which is an adaptively regularied CT image reconstruction can adaptively choose regularization parameters with making full use of the results of every iteration to update regularization parameters. Simulation results show that this method reduces the noise and improve image quality. The method is used to reconstruct the actual CT projection data to demonstrate the feasibility and practicality.OS-OSL-EM algorithm is applied to three-dimensional cone-beam image reconstruction and gets the better image.