{"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/convolutional-neural-networks-for-skull","title":"Convolutional Neural Networks for Skull-stripping in Brain MR Imaging using Consensus-based Silver standard Masks","arxiv_id":"1804.04988","date":"2018-04-13","proceeding":null,"authors":["Oeslle Lucena","Roberto Souza","Leticia Rittner","Richard Frayne","Roberto Lotufo"],"abstract":"Convolutional neural networks (CNN) for medical imaging are constrained by\nthe number of annotated data required in the training stage. Usually, manual\nannotation is considered to be the \"gold standard\". However, medical imaging\ndatasets that include expert manual segmentation are scarce as this step is\ntime-consuming, and therefore expensive. Moreover, single-rater manual\nannotation is most often used in data-driven approaches making the network\noptimal with respect to only that single expert. In this work, we propose a CNN\nfor brain extraction in magnetic resonance (MR) imaging, that is fully trained\nwith what we refer to as silver standard masks. Our method consists of 1)\ndeveloping a dataset with \"silver standard\" masks as input, and implementing\nboth 2) a tri-planar method using parallel 2D U-Net-based CNNs (referred to as\nCONSNet) and 3) an auto-context implementation of CONSNet. The term CONSNet\nrefers to our integrated approach, i.e., training with silver standard masks\nand using a 2D U-Net-based architecture. Our results showed that we\noutperformed (i.e., larger Dice coefficients) the current state-of-the-art SS\nmethods. Our use of silver standard masks reduced the cost of manual\nannotation, decreased inter-intra-rater variability, and avoided CNN\nsegmentation super-specialization towards one specific manual annotation\nguideline that can occur when gold standard masks are used. Moreover, the usage\nof silver standard masks greatly enlarges the volume of input annotated data\nbecause we can relatively easily generate labels for unlabeled data. In\naddition, our method has the advantage that, once trained, it takes only a few\nseconds to process a typical brain image volume using modern hardware, such as\na high-end graphics processing unit. In contrast, many of the other competitive\nmethods have processing times in the order of minutes.","url_abs":"http://arxiv.org/abs/1804.04988v1","url_pdf":"http://arxiv.org/pdf/1804.04988v1.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":"convolutional-neural-networks-for-skull","repo_url":"https://github.com/MICLab-Unicamp/CONSNet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"skull-stripping","task_name":"Skull Stripping"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}