A Review of Deep Learning CT Reconstruction From Incomplete Projection Data

Abstract

Computed tomography (CT) is a widely used imaging technique in both medical and industrial applications. However, accurate CT reconstruction requires complete projection data, while incomplete data can result in significant artifacts in the reconstructed images, compromising their reliability for subsequent detection and diagnosis. As a result, accurate CT reconstruction from incomplete projection data remains a challenging research area in radiology. With the rapid development of deep learning (DL) techniques, many DL-based methods have been proposed for CT reconstruction from incomplete projection data. However, there are limited comprehensive surveys that summarize recent advances in this field. This article provides a comprehensive overview of the current state-of-the-art DL-based CT reconstruction from incomplete projection data, including acrlong SV reconstruction, acrlong LA reconstruction, acrlong MAR, acrlong IT, and ring artifact reduction. This survey covers various DL-based solutions to the five problems, potential limitations of existing methods, and future research directions.