{"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/interpretable-lstms-for-whole-brain","title":"Analyzing Neuroimaging Data Through Recurrent Deep Learning Models","arxiv_id":"1810.09945","date":"2018-10-23","proceeding":null,"authors":["Armin W. Thomas","Hauke R. Heekeren","Klaus-Robert Müller","Wojciech Samek"],"abstract":"The application of deep learning (DL) models to neuroimaging data poses\nseveral challenges, due to the high dimensionality, low sample size and complex\ntemporo-spatial dependency structure of these datasets. Even further, DL models\nact as as black-box models, impeding insight into the association of cognitive\nstate and brain activity. To approach these challenges, we introduce the\nDeepLight framework, which utilizes long short-term memory (LSTM) based DL\nmodels to analyze whole-brain functional Magnetic Resonance Imaging (fMRI)\ndata. To decode a cognitive state (e.g., seeing the image of a house),\nDeepLight separates the fMRI volume into a sequence of axial brain slices,\nwhich is then sequentially processed by an LSTM. To maintain interpretability,\nDeepLight adapts the layer-wise relevance propagation (LRP) technique. Thereby,\ndecomposing its decoding decision into the contributions of the single input\nvoxels to this decision. Importantly, the decomposition is performed on the\nlevel of single fMRI volumes, enabling DeepLight to study the associations\nbetween cognitive state and brain activity on several levels of data\ngranularity, from the level of the group down to the level of single time\npoints. To demonstrate the versatility of DeepLight, we apply it to a large\nfMRI dataset of the Human Connectome Project. We show that DeepLight\noutperforms conventional approaches of uni- and multivariate fMRI analysis in\ndecoding the cognitive states and in identifying the physiologically\nappropriate brain regions associated with these states. We further demonstrate\nDeepLight's ability to study the fine-grained temporo-spatial variability of\nbrain activity over sequences of single fMRI samples.","url_abs":"http://arxiv.org/abs/1810.09945v2","url_pdf":"http://arxiv.org/pdf/1810.09945v2.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":"interpretable-lstms-for-whole-brain","repo_url":"https://github.com/ArrasL/LRP_for_LSTM","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"deep-learning","task_name":"Deep Learning"}],"methods":[{"method_slug":"lstm","method_name":"LSTM"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.09945","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}