Research On Reconfigurable X-ray Spectrum Detecting And Resolving Method For Spectral CT Imaging

Medical CT（computed tomography）is a noninvasive way to obtain images of human internal tissue and plays an important role in the early prevention and noninvasive treatment of the disease.Specifically,spectral CT has a number of advantages relative to conventional CT,especially in terms of material decomposition.It can depict the distributions of both the electron density and effective atomic number with projection data in two or more X-ray energy bins.For most of clinical and preclinical spectral CT systems,the effective energy bins are limited and pre-specified.However,complicated objects,with various chemical compositions,generally present different reconstruction quality at different energy bins.Immutable energy windows inevitably compromise the dynamic range of CT image quality.To obtain sufficient energy-sensitive information for the dual-energy CT imaging,in this thesis,we proposed a silicon-based pixel structure for X-ray spectral detection.The detector pixel was layered along the photon incident direction with the discrete anode electrodes.An X-ray beam was injected from one side of the pixel（edge-on geometry）,and a charge-packets readout method similar to a 3-phase charge coupled device is used.We investigate a reconfigurable energy-resolving method utilizing dynamic energy bins of a layered X-ray detector,and more spectral information captured in different energy bins can be obtained from a single CT scan.To accurately separate different energy bins,a rectification method can be used to reduce the overlap or cross-talk between high-and low-energy bins.To demonstrate the feasibility of the proposed methods,we simulate the photoelectric response of different dual-energy bins based on Edge-on detector pixel.Then,the images of effective atomic number Z_eff and electron density p_eff of representative phantoms are reconstructed from data in three pairs of dual-energy bins.The layered pixel structure gives an opportunity to obtain multiple energy-bin data by grouping/integrating charges of different layers.It is found that the average relative error between the theoretical and rectified charge counts is 7.31%,which indicates that the spectrum-rectification method can be used to estimate the charge counts generated by high-energy tail in shallower layers.The peak signal-to-noise ratio PSNRs of reconstruction images vary with the setting of energy bins and the size of phantoms.The flexibility in defining energy bins dynamically promises a higher reconstruction quality and a larger dynamic range of spectral CT.Our study gives an opportunity to optimize image quality by dynamically adjusting energy bins and reduce the radiation.