Study On Material Decomposition Algorithm Based On Image-domain For Spectral CT

X-ray CT utilizes the attenuation difference information caused by X-rays passing through the object to reconstruct a gray image reflecting the internal structure of the object.Traditional XCT has disadvantages such as low resolution and inability to identify and decompose materials.Spectral CT technique obtains projection data under different energy channels through photon counting detectors.It has the advantages of high resolution,high tissue contrast,low radiation dose,and material identification and decomposition.Material decomposition is an important application of spectral CT.How to improve the accuracy of material decomposition is one of the research hotspots of spectral CT.This thesis studies the material decomposition algorithms of spectral CT based on image domain,and proposes a material decomposition model based on multi constraint optimization,which is verified by simulation and actual experiments.The main work of this thesis is as follows:(1)Traditional reconstruction algorithm is used to reconstruct CT image,and the most basic direct decomposition method(DI)is used for decomposition by simulation and actual data.The results show that the decomposition accuracy is too low,noise is large,and details are not clear.(2)In order to overcome the defects of DI,the total variation based material decomposition model(TVMD)is studied based on the regularization theory.The solving process is deduced in detail,and the material decomposition experiments of simulation and actual data are carried out.Compared with DI,the quality of decomposition results is improved to a certain extent,the decomposition accuracy is improved,and the noise is suppressed to a certain extent.(3)In order to further improve the accuracy of material decomposition and image details,and overcome the limitations of second-order model,a material decomposition model(ROF-LLTMD)combining ROF model(TV)and fourth-order LLT model is studied.Solving process is derived in detail and experimental research is carried out.The results show that the image noise is suppressed obviously and the image is clearer,but the decomposition accuracy of contrast medium solution is still low.(4)In order to improve the resolution accuracy of contrast medium solution and improve the overall image quality,a multi constrained nonlocal total variation material decomposition model(MCNLTV)based on non local total variation(NLTV)and multi constraint is proposed.The air pixel processing method is introduced.Iterative solution process of the model was deduced in detail.The proposed algorithm is used for simulation and real data experiments.Compared with other methods,MCNLTV algorithm overcomes the defect of low resolution of contrast medium solution,and the obtained material composition diagram has better numerical evaluation while maintaining the overall quality.