A Study On Material Decomposition Methods Based On Spectral CT

Computed Tomography(CT)is an essential detection method in the field of medicine,and it plays an important role in the diagnosis of many diseases.Compared with traditional X-ray CT,spectral CT can obtain more material information by acquiring projection data in different energy bins,and has the ability of material identification and quantitative analysis.In recent years,more and more researchers have paid much attention to the research field.At present,the most widely used spectral CT is dual-energy CT in clinical application.With the rapid development of photon counting detectors,Spectral CT based on photon-counting detectors has shown the irreplaceable advantages over dual-energy CT.In order to achieve high-performance material decomposition algorithms,improve the accuracy of material decomposition and the quality of the decomposition coefficient image,this paper studies three material decomposition methods for both dual-energy CT and spectral CT based on photon counting detectors: image-domain material decomposition method,projection-domain material decomposition method and direct iterative material decomposition method.In terms of image-domain material decomposition,an image-domain material decomposition method based on the Maximum A Posteriori-Expectation Maximum(MAP-EM)algorithm is proposed.In this method,image reconstruction from projection is carried out firstly,and then material decomposition is implemented on the reconstructed image.Compared with the Filter Back-Projection(FBP)algorithm,the MAP-EM algorithm is able to reduce the noise level of the reconstructed image by the regularization.The experimental results of dual-energy CT show that compared with the image-domain material decomposition method based on FBP algorithm,the contrast noise ratio(CNR)of the reconstructed images is increased by 13.83-23.55%,and the error of the material decomposition coefficients is decreased by 32.50-90.40%.Furthermore,the image-domain material decomposition method based on the MAP-EM algorithm is applied to the spectral CT based on photon counting detectors.The experimental results show that compared with dual energy CT,for the human equivalent tissue materials,the error of the material decomposition coefficient is reduced by 14.14-58.32% in the spectral CT based on photon counting detectors.In terms of projection-domain material decomposition,considering that the projection-domain material decomposition method can effectively eliminate the problem of beam-hardening artifacts,a convenient energy spectral CT numerical simulation method is proposed and compared with the image-domain material decomposition methods.The experimental results of dual-energy CT show that for the equivalent material of human tissues,the noise level of decomposition coefficient images is reduced by 37.01-91.46% by the projection-domain material decomposition method,and the error of the decomposition coefficients is decreased by 1.22-25.10%.Furthermore,the projection-domain material decomposition method is applied to the spectral CT based on the photon counting detector and compared with the decomposition results of the dual energy CT.The experimental results show that the spectral CT based on the photon counting detector achieves better decomposition results,and the error level of the decomposition coefficient is decreased by 19.00-46.81%.In terms of direct iterative matter decomposition,a direct iterative projection function model is proposed,and then the MAP-EM algorithm is employed to solve it.The experimental results of dual-energy CT show that compared with the image-domain material decomposition method based on the FBP algorithm,the noise level of the decomposition coefficient image,which is obtained by the direct iterative material decomposition method,is reduced by 57.42-63.64%,and the error is decreased by 31.72-62.14%.Furthermore,a direct iterative material decomposition method in polar coordinate system is proposed.The experimental results show that the error of the decomposition coefficient of the direct iterative material decomposition method in Cartesian system is reduced by 1.86-29.35%,and the calculation speed is increased by 12.09% at the same time.Finally,the results of the image-domain material decomposition method based on MAP-EM algorithm,the projection-domain material decomposition method and the direct iterative material decomposition method in Cartesian and polar coordinate system are compared.The experimental result shows that the lowest error of the decomposition coefficients is obtained by the direct iterative material decomposition method in the polar coordinate system.