PSDP: Pseudo-supervised dual-processing for low-dose cone-beam computed tomography reconstruction

Lianying Chao, Wenqi Shan, Yanli Wang, Wenting Xu, Haobo Zhang, Zhiwei Wang, Qiang Li

Abstract: Highlights•A novel unpaired learning-based strategy for low-dose CT reconstruction.•A GAN-based DeProjNet for building clean-noisy training pairs.•A global-path and local-path discriminator for learning noise distribution.•A physics-based content-fidelity loss for avoiding the content deformation.