{"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-domain-guided-cnn-architecture-for","title":"A Domain Guided CNN Architecture for Predicting Age from Structural Brain Images","arxiv_id":"1808.04362","date":"2018-08-11","proceeding":null,"authors":["Pascal Sturmfels","Saige Rutherford","Mike Angstadt","Mark Peterson","Chandra Sripada","Jenna Wiens"],"abstract":"Given the wide success of convolutional neural networks (CNNs) applied to\nnatural images, researchers have begun to apply them to neuroimaging data. To\ndate, however, exploration of novel CNN architectures tailored to neuroimaging\ndata has been limited. Several recent works fail to leverage the 3D structure\nof the brain, instead treating the brain as a set of independent 2D slices.\nApproaches that do utilize 3D convolutions rely on architectures developed for\nobject recognition tasks in natural 2D images. Such architectures make\nassumptions about the input that may not hold for neuroimaging. For example,\nexisting architectures assume that patterns in the brain exhibit translation\ninvariance. However, a pattern in the brain may have different meaning\ndepending on where in the brain it is located. There is a need to explore novel\narchitectures that are tailored to brain images. We present two simple\nmodifications to existing CNN architectures based on brain image structure.\nApplied to the task of brain age prediction, our network achieves a mean\nabsolute error (MAE) of 1.4 years and trains 30% faster than a CNN baseline\nthat achieves a MAE of 1.6 years. Our results suggest that lessons learned from\ndeveloping models on natural images may not directly transfer to neuroimaging\ntasks. Instead, there remains a large space of unexplored questions regarding\nmodel development in this area, whose answers may differ from conventional\nwisdom.","url_abs":"http://arxiv.org/abs/1808.04362v1","url_pdf":"http://arxiv.org/pdf/1808.04362v1.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-domain-guided-cnn-architecture-for","repo_url":"https://github.com/saigerutherford/anatomically_defined_CNNs","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"object-recognition","task_name":"Object Recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}