Non-local and motion-based low-rank regularizations for gated CT reconstruction

Respiratory gated computed tomography (CT) is used for a diagnosis of certain diseases with motion patterns that causes motion blurring artifacts; and each phase is acquired under a low-dose condition that results in high noise in images. In this paper, to address the issues of high noise and motion artifacts in the gated CT, we propose an iterative reconstruction algorithm using 4D non-local and motion-based low-rank regularizations, which utilizes small image patches to reduce noise while preserving the edge structures, and exploits the low-rank property of the redundant pseudo motion information in multi-frame images. During the calculation of the non-local weights between inter-frame images, each voxel can achieve temporal positions of the highest non-local weights along frames, and we can generate a pseudo motion registration matrix for a specific phase image. Then, we can calculate the low-rank regularization using the pseudo motion registration matrix. The proposed method is computationally efficient in terms of the concurrent execution of calculating the non-local regularization and generating the pseudo registration matrix. In computer simulation, we demonstrate that the proposed method improves the image quality.