Spectral CT Reconstruction Algorithm Based On Energy Channel Correlation

Spectral CT based on photon counting detectors has great potential for applications such as material decomposition,tissue characterization,and lesion detection.Spectral CT can effectively avoid the influence of electronic noise through photon counting detectors,providing a higher signal-to-noise ratio than conventional CT detectors;Simultaneous acquisition of projection data under multiple energy channels of an object provides more image information than conventional CT,and excels in improving spatial resolution and low-dose imaging,which has become a hot spot in recent CT research.However,the obtained projection data are subject to noise interferences due to charge sharing,pulse effects and photon numbers,which pose a great challenge to its application.Therefore,obtaining higher quality reconstructed images from projection data is of great importance for practical applications.In this paper,we study the spectral CT reconstruction algorithm model from the perspective of reconstruction algorithm,considering the structural correlation and non-local feature similarity of multi-channel reconstructed images,and the main research work is as follows.（1）An Spectral CT reconstruction algorithm based on unclear norm of multi-channel joint total generalized variables（TGV）was proposed.The generalized total variables are extended to vectors to exploit the sparsity of singular values to promote linear dependence of image gradients and better recovery of image edge features.Simulation experiments with simple models and practical experiments with clinical mice validate the effectiveness of the algorithm.（2）a spectral CT reconstruction algorithm based on joint multi-channel TGV minimization and tensor decomposition（TD）was proposed.To better recover the image edges,the multi-channel joint TGV function is used as regularization.At the same time,it made full use of the similarity of non-local features in the image domain,clustered similar image blocks into fourth-order tensor groups,and suppressed noise in the reconstructed images by performing a sparse representation of the high-dimensional tensor.Experimental results show that the algorithm can further improve the quality of the reconstructed images while maintaining the edges and details of the energy-spectral CT images.Experimental results show that the algorithm can further improve the quality of the reconstructed images while maintaining the edges and details.