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

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 trains and inferences quickly, and the GAN promotes realistic output hardly distinguishable from original HR images. Our results from experiments on a dataset with 1,113 subjects shows that our new architecture outperforms other popular deep learning methods in recovering 4x resolutiondowngraded images and runs 6x faster.

Enhanced Discriminative Generative Adversarial Network for Face Super-Resolution

“…Note that in each RRDB the feature is residual summed; therefore, the overall number of features n f is unchanged throughout all blocks. Compared to mDCSRN proposed in [2], which densely accumulates features globally and grows wider as the network gets deeper, the RRDG remains narrow. This memory efficient aspect of RRDB is an important advantage for 3D applications as in the current brain MRI SISR task.…”

“…Ledig et al [9] introduced a residual network [5] for SISR and Zhang et al [19] extends the idea with the residual in residual connection. Chen et al [2], on the other hand, adapted the dense net to for the SISR task. Combining the residual connection and dense connection, Wang et al [18] proposed a hybrid of residual and dense connections, termed residual-in-residual dense block (RRDB), to replace the basic residual block in SRResNet.…”

“…The developments here applies generally to medical imagery but, for concreteness, are instantiated for SISR in T1 MR images of the brain, as in [2]. Assuming a k-space subsampled low resolution image as input, we propose to extend the GAN model for SISR as follows: 1) a new generator backbone is developed based on 3D residual-in-residual dense block [18], that is both high performance and memory efficient; 2) a patch GAN [8] discriminator that is more numerically stable; 3) an efficient blending approach is implemented to trade off between PSNR and GAN oriented models to generate the final output; and 4) a novel metric quantifying anatomical fidelity is introduced using a pre-trained brain parcellation network.…”

Virtual Thin Slice: 3D Conditional GAN-based Super-Resolution for CT Slice Interval

“…In medical imaging, few researchers work on the 3D image super resolution. Chen et al targeted 3D MRI super-resolution for medical image analysis [2]. Their target was limited to brain images.…”

Deep Learning Within a Priori Temporal Feature Spaces for Large-Scale Dynamic MR Image Reconstruction: Application to 5-D Cardiac MR Multitasking

“…We evaluate this approach for 5-D cardiac MR Multitasking, a low-rank tensor approach with three time-dimensions (cardiac phase, respiratory phase, and inversion time) and >40,000 frames per image sequence, enabling non-ECG, free-breathing, myocardial T1 mapping [12]. For this evaluation, we compare mDCN [13] and the stateof-the-art DenseUnet [14] as example network architectures, training on 153 subjects. We further expanded the mDCN structure with dilated convolutional layers for a larger receptive field, which is proven to further improve the reconstruction performance.…”