Research On X-ray Spectral CT Imaging And Component Characterization Method Based On Multispectral Projection Separation

X-ray spectral computed tomography(CT)is an effective method to suppress X-ray polychromatic hardening artifacts and improve the contrast of CT images,and is of great significance in the fields of new materials,new drugs,micro-nano devices,and so on.The existing spectral CT reconstruction methods are mainly based on iterative reconstruction algorithm.They are coupled with the physical prior of X-ray multispectral imaging,and are achieved by constructing a multispectral projection observation model and designing an iterative optimization solution algorithm.However,this type of algorithm is based on complex models,many assumptions and uncertain optimization parameters.It is difficult to be implemented in practical and engineering applications.Therefore,research work based on the decomposition of the projection domain to obtain narrow-energy-width projections has gradually been emphasized.The decomposition method based on the projection domain can directly obtain the projection data with good narrow spectrum characteristics,and then the CT image with great gray consistency can be obtained by traditional CT reconstruction algorithm.The decomposition method effectively reduces the design difficulty of CT reconstruction algorithm and provides a new multispectral CT imaging approach.Therefore,on the basis of studying the existing decomposition method with respect to the projection domain,the paper studies the X-ray multispectral CT imaging and component characterization method based on multispectral projection separation to improve CT image contrast and satisfy the needs of component differentiation and characterization.Based on the study of the principle of X-ray multispectral CT imaging,the paper introduces the base effect decomposition to solve the problem of no energy constraint about the multi-energy projection sequence blind separation model.According to the photoelectric effect and the Compton effect,the attenuation coefficient in the separation model is decomposed into energy and material terms,and a blind separation model of multispectral projection sequence based on the base effect decomposition is established.Simultaneously,by adopting the multispectral characteristics of radiography and the coupled residualnon-negative constraints,an iterative optimization algorithm for the decomposition model is derived.A sample composed of magnesium and aluminum was used as an object for experimental analysis.The experiment shows that the multispectral CT image sequence corresponding to the decomposed projection sequence has the clear energy spectrum directivity,and the accuracy of attenuation coefficient characterized by the reconstructed image is effectively improved.Secondly,aiming at the problem of high-precision mechanical relocation existing in multiple variable energy CT scans,on the basis of analyzing the projection imaging model and the multispectral projection sequence blind separation model,a multispectral CT imaging method based on staged scanning is proposed.According to the base effect decomposition,a CT reconstructed image is decomposed into space-dependent item(base image)and energy-dependent item(energy coefficient),which is substituted into the multispectral projection sequence blind separation model.A blind separation model with the base image as the solution target is established,and an iterative optimization algorithm for the decomposition model is derived.Experiments show that the method reduces the dependence on high relocation accuracy and the complexity of data acquisition and analysis while achieving near-narrow spectrum imaging.The requirement for multiple imaging of the same geometric parameter projection is avoided,and the scanning and calculation efficiency is improved.Finally,aiming at the problem of component characterization of multispectral CT sequence separated by the projection sequence,the paper studies the dual-energy CT characterization method based on image domain decomposition and analyzes the gray change trend of each material component in different separated CT sequences.On this basis,a multi-energy CT characterization model based on image domain decomposition is constructed.The proportion of each component in the CT image is solved by least square optimization,so as to realize the characterization and identification of material components.Experiments show that,compared with traditional dual-energy CT characterization methods,the proposed characterization method can effectively reduce the influence of noise caused by multispectral artifacts on component characterization and identification.Especially it has the better resolution ability for aluminum and silicon with similar attenuation characteristics.In addition,good experimental results have been obtained in the experiments of multi-material components.