MR image reconstruction methods from subsampled data using deep learning

Abstract

A long imaging time and metal-induced artifacts are major issues of MRI. In this paper, we applied deep learning to overcome these issues. Deep neural networks are proposed to solve these issues, and accompanying problems like insufficient training data, overfitting, and optimization are solved and analyzed.

To accelerate long imaging time, we propose a deep neural network to map aliased input images into desired alias-free images. The input of the deep neural network is all voxels in the aliased lines of multi-channel real and imaginary images from the subsampled k-space data, and the desired output is all voxels in the corresponding alias-free line of the root-sum-of-squares of multi-channel images from fully-sampled k-space data. Aliasing artifacts in an image reconstructed from subsampled data were reduced by line-by-line processing of the learned MLP architecture.

To reduce metal-induced artifacts, we propose a deep neural network based on supervised learning. The neural network is trained to map two distorted images obtained by dual-polarity readout gradients into a distortion-free image obtained by fully phase encoding. Simulated data were utilized to substitute real MR data for training.

In addition, an unsupervised learning method is proposed for a deep neural network architecture, consisting of a deep neural network and an MR image generation module. The architecture is trained as an end-to-end process without the use of distortion-free images or off-resonance frequency maps. The deep neural network estimates frequency-shift maps between two distorted images that are obtained using dual-polarity readout gradients. From the estimated frequency-shift maps and two distorted input images, distortion-corrected images are obtained with the MR image generation module.

Phantom and in vivo experiments were performed to compare the quality of reconstructed images. The proposed methods outperform the compared methods in quantitative and qualitative evaluations.