{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/deep-generative-adversarial-networks-for","title":"Deep Generative Adversarial Networks for Compressed Sensing Automates MRI","arxiv_id":"1706.00051","date":"2017-05-31","proceeding":null,"authors":["Morteza Mardani","Enhao Gong","Joseph Y. Cheng","Shreyas Vasanawala","Greg Zaharchuk","Marcus Alley","Neil Thakur","Song Han","William Dally","John M. Pauly","Lei Xing"],"abstract":"Magnetic resonance image (MRI) reconstruction is a severely ill-posed linear\ninverse task demanding time and resource intensive computations that can\nsubstantially trade off {\\it accuracy} for {\\it speed} in real-time imaging. In\naddition, state-of-the-art compressed sensing (CS) analytics are not cognizant\nof the image {\\it diagnostic quality}. To cope with these challenges we put\nforth a novel CS framework that permeates benefits from generative adversarial\nnetworks (GAN) to train a (low-dimensional) manifold of diagnostic-quality MR\nimages from historical patients. Leveraging a mixture of least-squares (LS)\nGANs and pixel-wise $\\ell_1$ cost, a deep residual network with skip\nconnections is trained as the generator that learns to remove the {\\it\naliasing} artifacts by projecting onto the manifold. LSGAN learns the texture\ndetails, while $\\ell_1$ controls the high-frequency noise. A multilayer\nconvolutional neural network is then jointly trained based on diagnostic\nquality images to discriminate the projection quality. The test phase performs\nfeed-forward propagation over the generator network that demands a very low\ncomputational overhead. Extensive evaluations are performed on a large\ncontrast-enhanced MR dataset of pediatric patients. In particular, images rated\nbased on expert radiologists corroborate that GANCS retrieves high contrast\nimages with detailed texture relative to conventional CS, and pixel-wise\nschemes. In addition, it offers reconstruction under a few milliseconds, two\norders of magnitude faster than state-of-the-art CS-MRI schemes.","url_abs":"http://arxiv.org/abs/1706.00051v1","url_pdf":"http://arxiv.org/pdf/1706.00051v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"deep-generative-adversarial-networks-for","repo_url":"https://github.com/bencottier/GANCS","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"deep-generative-adversarial-networks-for","repo_url":"https://github.com/gongenhao/GANCS","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"diagnostic","task_name":"Diagnostic"},{"task_slug":"mri-reconstruction","task_name":"MRI Reconstruction"},{"task_slug":"compressed-sensing","task_name":"compressed sensing"}],"methods":[{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"gan-least-squares-loss","method_name":"GAN Least Squares Loss"},{"method_slug":"lsgan","method_name":"LSGAN"},{"method_slug":"relu","method_name":"ReLU"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1706.00051","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}