Efficient and Accurate MRI Super-Resolution using a Generative Adversarial Network and 3D Multi-Level Densely Connected Network

TL;DR

This paper introduces a novel 3D GAN-guided neural network architecture for MRI super-resolution that achieves faster training and inference while producing highly realistic high-resolution images from low-resolution inputs.

Contribution

The paper presents a multi-level densely connected 3D neural network with GAN training for efficient and accurate MRI super-resolution, outperforming existing methods in speed and quality.

Findings

Outperforms other deep learning methods in image recovery quality.

Runs 6 times faster than comparable models.

Effective in 4x resolution enhancement for MRI images.

Abstract

High-resolution (HR) magnetic resonance images (MRI) provide detailed anatomical information important for clinical application and quantitative image analysis. However, HR MRI conventionally comes at the cost of longer scan time, smaller spatial coverage, and lower signal-to-noise ratio (SNR). Recent studies have shown that single image super-resolution (SISR), a technique to recover HR details from one single low-resolution (LR) input image, could provide high-quality image details with the help of advanced deep convolutional neural networks (CNN). However, deep neural networks consume memory heavily and run slowly, especially in 3D settings. In this paper, we propose a novel 3D neural network design, namely a multi-level densely connected super-resolution network (mDCSRN) with generative adversarial network (GAN)-guided training. The mDCSRN quickly trains and inferences and the GAN promotes realistic output hardly distinguishable from original HR images. Our results from experiments on a dataset with 1,113 subjects show that our new architecture beats other popular deep learning methods in recovering 4x resolution-downgraded im-ages and runs 6x faster.