A Research On Image Reconstruction Method For Multi-energy Photon Counting X-CT

Multi-energy photon counting X-CT(MPCCT) with high energy resolution makes the detected photon in energy-bins carry abundant material attenuation information, which has the ability to improve material identification and componential differentiation of images and has become a research focus in recent years. However, compared with the traditional imaging, MPCCT record the relatively fewer photons in each bin, thus the signal to noise ratio decrease, moreover, there may be much noise in projection datasets because of some deficiencies in current photon counting detector. Images reconstruction with analytical algorithm could meet the general application requirements, however, it is lack of the ability to restrain noise, which is difficult to be applied for MPCCT, image reconstruction with iterative algorithm could restrain noise in some extent, but the image detail information is hard to keep. Therefore, it is necessary to study the reconstruction method for MPCCT to effectively improve the image quality.For the image reconstruction of MPCCT in this paper, energy weighted image reconstruction method has been investigated based on analytic method, the abundant material attenuation information of transmitted spectrum is used to aquire the full spectrum image of high quality. Moreover, image reconstruction method based on prior image and compressed sensing(PICS) has been investigated to improve the image quality of all energy-bins effectively. Firstly, both projection-based and image-based weighted scheme are investigated for image reconstruction respectively, moreover, for projection-based weighted scheme, noise information directly extracted from projection sine diagram that contains multiple kinds of noise has been proposed to derive weight for image reconstruction. Secondly, the generated PI has been taken as constraint condition and combined with CS to derive objective function, and then iterative method has been used to reconstruct image, the main factor of affecting image reconstruction has been deeply analyzed.To evaluate the validity of the proposed method for improving image quality, numerical phantoms designed by Genat4 have been reconstructed and analyzed. Experimental results show that energy weighted image reconstruction method can increase contrast-to-noise ratio(CNR) of reconstructed full spectrum image, and the higher the contrast between the materials, the better the CNR of reconstructed image; Image reconstructed from the image-based weighted scheme is better than that from the projection-based weighted scheme; To some extent, noise information directly extracted from projection sine diagram could be very useful for improving quality with projection-based weighted scheme. The reconstruction method based on PI and CS could greatly improve the image quality. Compared with FBP, it is able to reduce noise in reconstructed images of three energy bins by 78.96%, 82.16%, 88.91% respectively, meanwhile, the root-mean-square error(RMSE) in each energy bin is decreased by 14.7%, 18.25%, 34.51% and the correlation coefficient(CC) is raised by 9.63%, 11.24%, 15.94% respectively; Moreover, optimization method for objective function, weight scheme for obtaining prior image, prior image parameter ? and energy threshold could affect image quality. The results suggest that it could improve image quality of all energy-bins using separable paraboloidal surrogates(SPS) algorithm to optimize objective function, reconstructing prior image from image-based weighted scheme, choosing suitable prior image parameter, and setting energy threshold according to transmission spectrum.