Deep Generative Adversarial Networks for Compressed Sensing Automates MRI

Abstract

Magnetic resonance image (MRI) reconstruction is a severely ill-posed linear inverse task demanding time and resource intensive computations that can substantially trade off {\it accuracy} for {\it speed} in real-time imaging. In addition, state-of-the-art compressed sensing (CS) analytics are not cognizant of the image {\it diagnostic quality}. To cope with these challenges we put forth a novel CS framework that permeates benefits from generative adversarial networks (GAN) to train a (low-dimensional) manifold of diagnostic-quality MR images from historical patients. Leveraging a mixture of least-squares (LS) GANs and pixel-wise ${\ell }_1$ cost, a deep residual network with skip connections is trained as the generator that learns to remove the {\it aliasing} artifacts by projecting onto the manifold. LSGAN learns the texture details, while ${\ell }_1$ controls the high-frequency noise. A multilayer convolutional neural network is then jointly trained based on diagnostic quality images to discriminate the projection quality. The test phase performs feed-forward propagation over the generator network that demands a very low computational overhead. Extensive evaluations are performed on a large contrast-enhanced MR dataset of pediatric patients. In particular, images rated based on expert radiologists corroborate that GANCS retrieves high contrast images with detailed texture relative to conventional CS, and pixel-wise schemes. In addition, it offers reconstruction under a few milliseconds, two orders of magnitude faster than state-of-the-art CS-MRI schemes.

References in corpus (7)

• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Low-Dose CT with a Residual Encoder-Decoder Convolutional Neural Network (RED-CNN)
• Perceptual Losses for Real-Time Style Transfer and Super-Resolution
• Identity Mappings in Deep Residual Networks
• Least Squares Generative Adversarial Networks
• Generative Visual Manipulation on the Natural Image Manifold
• Real-time Dynamic MRI Reconstruction using Stacked Denoising Autoencoder

Cited by in corpus (38)

• Generative Adversarial Network in Medical Imaging: A Review
• FastSurfer -- A fast and accurate deep learning based neuroimaging pipeline
• Compressed Sensing MRI Reconstruction using a Generative Adversarial Network with a Cyclic Loss
• How Generative Adversarial Networks and Their Variants Work: An Overview
• Real-time Cardiovascular MR with Spatio-temporal Artifact Suppression using Deep Learning - Proof of Concept in Congenital Heart Disease
• Compressed Sensing with Deep Image Prior and Learned Regularization
• Motion Corrected Multishot MRI Reconstruction Using Generative Networks with Sensitivity Encoding
• Solving inverse problems using conditional invertible neural networks
• Data synthesis and adversarial networks: A review and meta-analysis in cancer imaging
• Deep Bayesian Inversion
• High-Resolution Limited-Angle Phase Tomography of Dense Layered Objects Using Deep Neural Networks
• Leveraging Uncertainty Estimates for Predicting Segmentation Quality
• Low-Dose CT with Deep Learning Regularization via Proximal Forward Backward Splitting
• Robust Compressed Sensing MRI with Deep Generative Priors
• Recurrent Generative Adversarial Networks for Proximal Learning and Automated Compressive Image Recovery
• Task adapted reconstruction for inverse problems
• Highly Scalable Image Reconstruction using Deep Neural Networks with Bandpass Filtering
• ReconResNet: Regularised Residual Learning for MR Image Reconstruction of Undersampled Cartesian and Radial Data
• Learning to solve inverse problems using Wasserstein loss
• Limited-angle tomographic reconstruction of dense layered objects by dynamical machine learning
• Unified Supervised-Unsupervised (SUPER) Learning for X-ray CT Image Reconstruction
• Geometric Approaches to Increase the Expressivity of Deep Neural Networks for MR Reconstruction
• Generative adversarial networks and adversarial methods in biomedical image analysis
• A Provably Convergent Scheme for Compressive Sensing under Random Generative Priors
• Big in Japan: Regularizing networks for solving inverse problems
• Deep learning versus ${\ell }^1$-minimization for compressed sensing photoacoustic tomography
• Multi-institutional Collaborations for Improving Deep Learning-based Magnetic Resonance Image Reconstruction Using Federated Learning
• Deep Learning Techniques for Inverse Problems in Imaging
• Variational Inference for Computational Imaging Inverse Problems
• MRI Image Reconstruction via Learning Optimization Using Neural ODEs
• Deep Learning Techniques for Compressive Sensing-Based Reconstruction and Inference -- A Ubiquitous Systems Perspective
• Learn distributed GAN with Temporary Discriminators
• Deep Decomposition Learning for Inverse Imaging Problems
• Compressive Learning of Generative Networks
• Generative Model Adversarial Training for Deep Compressed Sensing
• Reconstructing unseen modalities and pathology with an efficient Recurrent Inference Machine
• Tree Search Network for Sparse Regression
• Unpaired Deep Learning for Accelerated MRI using Optimal Transport Driven CycleGAN