Research On Reconstruction Algorithms And Comprehensive Applications For Spectral CT

CT is an important equipment of nondestructive testing,which is widely implemented in many fields,such as medical examination and safety inspection.With the improvement of the nuclear electronic processing speed,the development of photon counting detectors(PCDs)is becoming more and more effective to the practical application,and more attention has been paid to the development of spectral CT.The PCDs are able to get multiple energy window data in one scan.Compared with traditional CT,it can provide material resolution and reduced noise.Compared with dual energy CT,it can effectively reduce dose and achieve quantitative K-edge imaging.With the aim at the optimal use of spectral information,the dissertation studied spectral CT on both reconstruction algorithms and the comprehensive applications.The response function of the detector is the basis for the optimal utilization of the spectral information.We first propose a physical model for the response function of the PCDs,which combined Monte Carlo simulation and empirical model.Geant4 is used to simulate the deposited energy spectrum of monochromatic photon.The response function of the deposited energy contains the Gauss function and the empirical function.X-ray fluorescence spectrum is used for calibration of the key parameters in the model.The experimental results show that our model is accurate.The inconsistency correction of the detector pixels consists of the threshold correction and the response function correction,which effectively suppressed the ring artifact in the reconstruction.The maximum-likelihood(ML)decomposition algorithm can reach the Cramer-Rao lower bound(CRLB).Based on the response function and physical model,we implemente the ML decomposition algorithm and verify it with multiple materials.The experimental results verify the accuracy and effectiveness of the algorithm.Based on the ML decomposition algorithm,we simulated the relationship between the number of energy bins and the noise,and the experiment results verify the simulation results.We calibrate the deviation of the system based on a piecewise fitting method,and deduce the decomposition bias based on the system deviation.The experiment results verify the conclusion.We propose an implementation of spectrum CT based liquid security inspection.We first estimated the scatter based on the unblocked pixels.Then,the ML decomposition algorithm with three energy channels is used to extract the photo-electric coefficient and Compton scattering coefficient.Experiments show that this system can extract accurate information of liquid inside a metal container.Compared with the traditional dual energy CT,this system has higher accuracy and stability.A set of image de-noising method is proposed.The noise model of the decomposition projections is introduced to the de-noising process based on TV.The quasi-monochromatic image is introduced to the reconstruction image denoising process based on the bilateral filter optimization method.The experimental results show that the spectral CT can reduce the hardening artifact,and both the spatial resolution and the density resolution are higher than the traditional dental CT.