SUMMARY

Compared to commercial dual-energy computed tomography (DECT) solutions, beam modulation is an efficient and cost-effective method to achieve single-scan DECT with conventional diagnostic or cone-beam CT machines. However, current beam modulation methods utilize regularization strategies for DECT reconstruction, which are unsuitable for half-fan and short-scan modes, resulting in data loss in each spectrum. Additionally, previous methods use a primary-modulation scheme which leads to a magnified penumbra area on the detector side. To address these limitations, a novel static detector modulation system is proposed in this dissertation to enable dual-energy projection acquisition during a single scan. The detector-modulated projection data is then processed using a pre-trained U-Net model with attention mechanism to generate full-sampled datasets at high- and low-energy spectra. The synthesized complete dual-energy data can then be used for DECT reconstruction and material decomposition.