Limited-angle computed tomography (CT) image reconstruction is a challenging problem\nin the field of CT imaging. In some special applications, limited by the geometric space and\nmechanical structure of the imaging system, projections can only be collected with a scanning range\nof less than 90°. We call this kind of serious limited-angle problem the ultra-limited-angle problem,\nwhich is difficult to effectively alleviate by traditional iterative reconstruction algorithms. With the\ndevelopment of deep learning, the generative adversarial network (GAN) performs well in image\ninpainting tasks and can add effective image information to restore missing parts of an image. In\nthis study, given the characteristic of GAN to generate missing information, the sinograminpainting-\nGAN (SI-GAN) is proposed to restore missing sinogram data to suppress the singularity\nof the truncated sinogram for ultra-limited-angle reconstruction. We propose the U-Net generator\nand patch-design discriminator in SI-GAN to make the network suitable for standard medical CT\nimages. Furthermore, we propose a joint projection domain and image domain loss function, in\nwhich the weighted image domain loss can be added by the back-projection operation. Then, by\ninputting a paired limited-angle/180° sinogram into the network for training, we can obtain the\ntrained model, which has extracted the continuity feature of sinogram data. Finally, the classic CT\nreconstruction method is used to reconstruct the images after obtaining the estimated sinograms.\nThe simulation studies and actual data experiments indicate that the proposed method performed\nwell to reduce the serious artifacts caused by ultra-limited-angle scanning.
Loading....