论文部分内容阅读
Purpose: To improve the image reconstruction quality of multi-energy photon counting X-CT based on prior image constrained compressed sensing by evaluation of several different optimization methods.Methods: A numerical phantom is designed to yield the projection datasets of three energy bins by using Geant4.Based on PICCS,the optimization objective function is constructed where the prior image is derived with energy weighted image reconstruction method.Several minimization methods including steepest descent(SD)algorithm,conjugate gradient(CG)algorithm and separable paraboloidal surrogates(SPS)algorithm are employed to solve the problem.The performances of the three optimization methods are assessed with respect to speed and stability of convergence and accuracy of reconstruction.The algorithm that provides the best balance of stability and convergence speed is used to study the performance of PICCS with respect to the prior image parameter α.The PICCS reconstruction method is evaluated in terms of reconstruction accuracy and noise characteristics of the reconstructed images of each energy bin.Results: SPS algorithm offers the best performance with respect to the iterative initial values,the minimum of objective functions.Using this minimization method,the overall image quality including the structural contour and image noise level achieves the best with prior image parameter α as 0.5.Compared with FBP,PICCS is able to reduce noise in reconstructed images by 80.46%、82.51%、88.08%in each energy bin respectively,meanwhile,the RMSE in each energy bin is decreased by 15.02%、18.15%、34.11%and CC is raised by 9.98%、11.38%、15.94%respectively.Conclusions: PICCS along with SPS algorithm is the best choice for the image reconstruction of multi-energy photon counting X-CT.