{"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/a-hybrid-frequency-domainimage-domain-deep","title":"A Hybrid Frequency-domain/Image-domain Deep Network for Magnetic Resonance Image Reconstruction","arxiv_id":"1810.12473","date":"2018-10-30","proceeding":null,"authors":["Roberto Souza","Richard Frayne"],"abstract":"Decreasing magnetic resonance (MR) image acquisition times can potentially\nreduce procedural cost and make MR examinations more accessible. Compressed\nsensing (CS)-based image reconstruction methods, for example, decrease MR\nacquisition time by reconstructing high-quality images from data that were\noriginally sampled at rates inferior to the Nyquist-Shannon sampling theorem.\nIn this work we propose a hybrid architecture that works both in the k-space\n(or frequency-domain) and the image (or spatial) domains. Our network is\ncomposed of a complex-valued residual U-net in the k-space domain, an inverse\nFast Fourier Transform (iFFT) operation, and a real-valued U-net in the image\ndomain. Our experiments demonstrated, using MR raw k-space data, that the\nproposed hybrid approach can potentially improve CS reconstruction compared to\ndeep-learning networks that operate only in the image domain. In this study we\ncompare our method with four previously published deep neural networks and\nexamine their ability to reconstruct images that are subsequently used to\ngenerate regional volume estimates. We evaluated undersampling ratios of 75%\nand 80%. Our technique was ranked second in the quantitative analysis, but\nqualitative analysis indicated that our reconstruction performed the best in\nhard to reconstruct regions, such as the cerebellum. All images reconstructed\nwith our method were successfully post-processed, and showed good volumetry\nagreement compared with the fully sampled reconstruction measures.","url_abs":"http://arxiv.org/abs/1810.12473v1","url_pdf":"http://arxiv.org/pdf/1810.12473v1.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":"a-hybrid-frequency-domainimage-domain-deep","repo_url":"https://github.com/rmsouza01/Hybrid-CS-Model-MRI","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"image-reconstruction","task_name":"Image Reconstruction"},{"task_slug":"compressed-sensing","task_name":"compressed sensing"}],"methods":[{"method_slug":"concatenated-skip-connection","method_name":"Concatenated Skip Connection"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"u-net","method_name":"U-Net"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}