Classification and Progress of Deep Learning-based MR Image Reconstruction

Abstract

The application of compressed sensing and related iterative reconstruction methods have been investigated to reduce the scan time of MRI. With the recent development of deep learning technology in the image processing field, deep learning-based image reconstruction methods have replaced iterative reconstruction methods as the mainstream method. Since the effectiveness of deep learning reconstruction methods has been well confirmed, researches are being expanded more widely and deeply to solve problems for practical applications and to introduce the latest technologies that are attracting attention in the field of natural image processing. In this paper, we categorize the deep learning image reconstruction methods proposed so far, and introduce recent trends and methods that have attracted much attention, showing the results of reconstruction experiments.