Research On Region Extraction And Fast Bayesian Matching Pursuit For X-ray Luminescence Computed Tomography

X-ray luminescence computed tomography(XLCT)is an important molecular imaging technology,which collects the NIR light emitted from the nanophosphors in biological applications that is irradiated by X-rays and combines reconstruction algorithm to obtain the three-dimensional(3-D)distribution of x-ray-excitable nanophosphors.XLCT contributes to monitor biological processes at the level of cells and molecules and becomes a powerful tools for the domain of early disease diagnosis as well as imaging-guided radiation therapy.Based on the X-ray and photon propagation model in vivo,the paper focus on the accuracy and speed of XLCT reconstruction,meanwhile,employ the effective algorithms to mitigate the XLCT ill-posedness.Combined high-efficiency XLCT reconstruction algorithm with the collected optical information that has been processing provides the theoretical and experimental support for the further development and practice of XLCT.The research and work in the paper can be summarized as follows:(1)The permissible region(PR)is used as a priori information for the reconstruction of XLCT which is helpful for the accuracy and effiency of reconstruction.However,the range of PR is mainly selected aritificially.The improper setting will affect the accuracy of results directly.Using the data of imaging object,we proposed a permissible region extraction strategy based on a knowledge priori N groups of recovered result with N groups of different discretized mesh have provided N groups of PR for XLCT,which can be considered as a knowledge priori,.The insertion of N groups of PR provides a reasonable and feasible PR of nanophosphor,and uses orthogonal matching pursuit(OMP)algorithm for XLCT reconstruction.A plenty of experiments results indicates that we obtained a best recovered results both in location and speed using the PR extraction strategy and OMP algorithm compared with the PR by double-mesh strategy as well as OMP algorithm without the proposed strategy.In addition,the experiment results with different noise also verifies the robustness of the proposed strategy.(2)Considering that multi-view imaging requires a longer scanning time,which is not helpful to image the target rapidly in vivo,we use a single-view x-ray luminescence computed tomography(XLCT)imaging to reconstruct the target.Single-view reconstruction suffers from a severe ill-posed problem compared with multi-view imaging as a result of only one angle data used in the reconstruction,based on the improvement of reconstruction quality and the reduction of scanning time,we present a fast Bayesian matching pursuit(FBMP)algorithm combined with iterative-shrinking permissible region(ISPR)strategy to implement a real-time XLCT imaging.In the method,bayesian model was combined with the greedy algorithm to quickly and efficiently restore sparse signal from few observed values.Meanwhile,FBMP was combined with ISPR strategy to simplify the mesh generation and system matrix construction by self-adaptive finite element,downsizing the permissible region iteratively with multiple reconstruction to reduce the instability of algorithm.A plenty of experiments results demonstrats that the proposed method produced an optimal result not only in the localization accuracy of nanophosphors but also the quantitative result of luminescence yield compared with traditional reconstruction algorithms,the efficiency is also improved significantly.Different optical parameters and noise were added in the reconstruction,the results further showed the stability of FBMP.