CT-Net: Cascaded T-shape network using spectral redundancy for dual-energy CT limited-angle reconstruction

• Published in: Biomedical signal processing and control Vol. 79; p. 104072
• Main Authors: Chen, Kai, Ji, Guohui, Wang, Chenrui, Peng, Zhiguang, Ji, Xu, Tang, Hui, Yang, Chunfeng, Chen, Yang
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.01.2023
• Subjects: Deep learning; Dual-energy computed tomography; Iodine contrast agent quantification; Limited-angle reconstruction; Virtual non-contrast (VNC) imaging; Deep learning; Virtual non-contrast (VNC) imaging; Iodine contrast agent quantification; Dual-energy computed tomography; Limited-angle reconstruction
• ISSN: 1746-8094; 1746-8108
• DOI: 10.1016/j.bspc.2022.104072

Dual-energy computed tomography (DECT) shows promising clinical significance in substance identification and quantitative analysis. Mostly dual-energy CT scanning systems use two sets of x-ray sources and detectors for full scanning to simultaneously acquire X-ray data of materials at high- and low-energy levels. The reconstructed high- and low-energy CT images have spectral redundancy in the energy domain. We propose a one-step dual-energy limited-angle reconstruction scheme exploiting the energy domain spectral redundancy. The scheme consists of a sinogram domain network(SD-Net), a reconstruction unit(RU), and an image domain network(ID-Net). After SD-Net complements the dual-energy limited-angle incomplete projection data, a RU is used to reconstruct the CT images. Finally, ID-Net processes the reconstructed CT images into high-quality CT images. We propose a Cascaded T-shape Network(CT-Net) based on spectral redundancy to improve DECT image quality. The CT-Net consists of a backbone net, a low-energy branch, and a high-energy branch. CT-Net can directly map incomplete projection data into high-quality DECT images that can be used for clinical diagnosis. Qualitative and quantitative results demonstrate the excellent performance of CT-Net in preserving edges, removing artifacts, and suppressing noise. Two common DECT applications, such as virtual non-contrast (VNC) imaging and iodine contrast agent quantification, prove the clinically promising potential of CT-Net. •We propose a one-step dual-energy limited-angle reconstruction strategy based on CT-Net.•We designed a novel loss function for dual-energy limited-angle reconstruction in CT-Net.•DE material decomposition was performed to validate the clinical utility.