Room: Track 1
Purpose: Photon-counting CT has been developed in past decades for various applications in medical imaging and material science. The recorded spectral inconsistency is still a major challenge that directly leads to artifacts in reconstructed CT images and to inaccuracies in material discrimination. The purpose of this study is to develop a physics model of spectral inconsistency and to achieve adaptive, fast and accurate spectral correction.
Methods: Combining with theoretical calculation and experimental data analysis, we hypothesize that the complicated spectral inconsistency can be expressed by two main factors: (1) pixel related energy threshold variations, and (2) spectral distortion at each pixel. Based on these assumptions, we establish a new spectral model for photon-counting detector (PCD) CT with the spectral inconsistency characterized by energy threshold variation factor and a spectral distortion factor, and proposed an adaptive spectral inconsistency correction approach. Both numerical simulations and physics experiments are carried out to validate our theory and method, on a tabletop PCD CT system that is composed of a VJP051 X-ray source and an XCounter FliteX1 CdTe PCD with 1536 pixels per row and pixel size being 0.1 mm × 0.1 mm.
Results: In simulations, modeled factors calculated using flat filters transmissions are in good agreement with their true values, with the conformity accuracy up to 99.999%. The image quality of CT reconstruction improved significantly after spectral inconsistency correction. In physics experiments, the conformity between measured value and calculated value is greater than 97.8%, artifacts were also reduced. All these results prove the effectiveness and feasibility of our proposed model.
Conclusion: Based on our preliminary results, the spectral inconsistency can be effectively and adaptively modeled and can be used to reduce related artifacts in CT reconstruction. Further optimization is under way to improve its performance on real PCD CT system.
Funding Support, Disclosures, and Conflict of Interest: This project was supported in part by the New Faculty Startup Funding of Tsinghua University, Beijing, China (53331100120).