Study On Material Decomposition Method Based On Spectral CT

With the rapid development of photon counting detection technology,X-ray multi-energy spectral CT(spectral CT)imaging technology has become a research hotspot in the field of CT imaging.Spectral CT systems based on photon counting detectors can simultaneously obtain projection data of multiple energy channels in one scan,and have great potential in applications such as material decomposition and tissue characterization.In this paper,based on the spectral CT system of photon counting detector,the material decomposition algorithm is studied.The main research work is as follows:(1)Multi-energy spectral CT images were analyzed based on principal component analysis.Principal component analysis(PCA)was used to analyze the spectral CT projection domain and image domain data.At the same time,a method combining dual-domain filtering and pixel value square was proposed to remove the background noise in the principal component image,and the principal component image in the image domain was compared systematically with the reconstructed image of the principal component in the projection domain,and then the selected principal component images were fused in color to obtain the color representation image of the material in the projection domain and the image domain.The experimental results show that compared with PCA in the image domain,the image details of PCA in the projection domain are more obvious,and the material identification is more accurate.(2)A projection domain material decomposition algorithm based on the combination of specific material regularization terms and total generalized variation(SRTGV)was proposed.This algorithm decomposes multiple materials in parallel,using specific regularization terms for each material,which can improve the decomposition accuracy of each material.Adding total generalized variation(TGV)regularization terms can not only reduce the impact of noise on material decomposition,but also protect the edges of the image.Experimental results show that the material decomposition method proposed in this paper has higher resolution and clearer images compared to the least square method and the decomposition method based on specific material regularization terms.