OASIS (dataset)

OASIS (dataset)
Name	OASIS (dataset)

OASIS (dataset) OASIS (dataset) is a neuroimaging dataset widely used in computational neuroscience, medical imaging, and machine learning research. It provides structural magnetic resonance imaging data and metadata for studies related to aging, dementia, and brain morphology, and has been cited across research associated with institutions such as Massachusetts Institute of Technology, Harvard University, Johns Hopkins University, Stanford University, and University of California, Berkeley. The dataset has informed analyses in projects connected to Alzheimer's Disease Neuroimaging Initiative, Human Connectome Project, OpenNeuro, UK Biobank, and research groups at Carnegie Mellon University.

Overview

OASIS (dataset) comprises cross-sectional and longitudinal structural MRI scans from cohorts covering healthy aging and cognitive impairment, enabling comparisons across populations studied at Massachusetts General Hospital, Brigham and Women's Hospital, University of Pennsylvania, Columbia University, and Yale University. The resource has been integrated into pipelines developed at National Institutes of Health, European Union research consortia, Wellcome Trust–funded projects, and by teams at University College London and McGill University. Researchers publishing in venues like Nature Medicine, Neuron, Proceedings of the National Academy of Sciences, IEEE Transactions on Medical Imaging, and Journal of Neuroscience have leveraged the dataset for morphometric analyses, comparative studies with Alzheimer's Disease Neuroimaging Initiative cohorts, and validation against data from ADNI partners and repositories such as NeuroVault.

Dataset Composition and Acquisition

The dataset contains T1-weighted MRI volumes obtained using scanners from manufacturers represented by Siemens AG, GE Healthcare, and Philips at field strengths associated with projects at Cleveland Clinic, Mayo Clinic, and Mount Sinai Health System. Subject recruitment protocols were informed by research ethics boards at Boston University, Brown University, Dartmouth College, and University of Michigan, with demographic and clinical assessments aligned to standards used by Alzheimer's Disease Neuroimaging Initiative, Framingham Heart Study, Rotterdam Study, and datasets curated at King's College London. Acquisition parameters reflect sequences cited in method sections of papers from Cambridge University, Imperial College London, and University of Oxford research teams, allowing cross-study harmonization with resources like ENIGMA and preprocessing approaches developed at Allen Institute for Brain Science.

Preprocessing and Annotation

Preprocessing workflows applied to the dataset often mirror pipelines from tools developed by FMRIB, FreeSurfer, SPM, AFNI, and ANTS, and adopt skull-stripping, bias-field correction, and registration procedures used in analyses from University of Southern California, Duke University, and University of Pittsburgh. Manual and automated annotations include tissue segmentations, cortical thickness maps, and volumetric labels that align with atlases such as the MNI (Montreal Neurological Institute) template, Harvard-Oxford Atlas, Desikan–Killiany atlas, and parcellations used in studies at Max Planck Society and Institut Pasteur. Quality control protocols draw on practices described in publications by European Organization for Nuclear Research, National Institute of Mental Health, National Institute on Aging, and groups at Riken, with metadata linking to cognitive assessments similar to those from Mini-Mental State Examination studies and clinical scales used in trials at Johns Hopkins Medicine.

Access, Licensing, and Usage

Access to the dataset is governed by data use agreements and ethical oversight mechanisms comparable to procedures at National Institutes of Health, Institutional Review Board, Wellcome Trust, and consortia like Global Alliance for Genomics and Health. Licensing terms permit research use in academic projects at Princeton University, Cornell University, University of Toronto, and University of British Columbia, while derivative datasets have been redistributed via platforms like OpenNeuro, Figshare, and repositories maintained by Dryad and Zenodo. Users often cite compliance workflows modeled on guidance from WHO, FDA, European Medicines Agency, and data governance frameworks developed at Carnegie Institution for Science and Salk Institute.

Evaluation Protocols and Benchmarks

Benchmarks using the dataset have been reported in comparative studies involving algorithms from groups at Google Research, Facebook AI Research, Microsoft Research, DeepMind, and academic teams at University of Toronto. Evaluation metrics include segmentation Dice coefficients, cortical thickness correlations, and classification accuracy for diagnostic labels, tested in competitions hosted by MICCAI, ISBI, NeurIPS, and workshops at CVPR. Cross-validation strategies reference statistical methods popularized in publications from Stanford University School of Medicine, Princeton Neuroscience Institute, and ETH Zurich, with reproducibility efforts coordinated alongside initiatives at Open Science Framework, Center for Open Science, and Berkeley Institute for Data Science.

Applications and Impact

The dataset has supported studies of brain aging, biomarker discovery, machine learning model development, and translational research at institutions like UCLA, University of Washington, University of California, San Diego, and Vanderbilt University Medical Center. It has informed clinical research related to neurodegenerative disorders investigated at Alzheimer's Association conferences, influenced normative modeling work in labs at Columbia University Irving Medical Center, and been used for training deep learning architectures pioneered at MIT CSAIL and evaluated by teams at KTH Royal Institute of Technology and Technical University of Munich. Ongoing impact includes integration into educational materials at Coursera, edX, and university curricula at Johns Hopkins Bloomberg School of Public Health, promoting open science practices championed by Plan S and publishing initiatives at Nature Publishing Group.

Category:Neuroimaging datasets