PhysioNet
Overview
PhysioNet is a major open source of electrophysiological signal data for neuroscience, hosting EDF-format EEG and polysomnography (sleep) recordings; its annual George B. Moody Challenge has run neuroscience tasks including EEG-based sleep staging and the prediction of cognitive impairment from sleep studies. Maintained by the MIT Laboratory for Computational Physiology and supported by the NIH since 1999, it is an open repository for physiological signal and clinical data, with deposit, versioning, persistent citation, and tiered access. Alongside its neurophysiology holdings it is the home of large de-identified clinical datasets, most prominently the MIMIC critical-care database.
Data and software
PhysioNet hosts biomedical signal and clinical datasets across modalities including EEG, ECG, polysomnography, and vital-signs waveforms, as well as imaging-linked clinical data (the MIMIC-CXR chest-radiograph database is stored in DICOM). Its native signal format is WFDB (WaveForm DataBase), maintained as an open software library that remains widely used for ECG and physiological signal processing; EEG and sleep recordings are commonly held in EDF. PhysioNet datasets are frequently converted downstream into BIDS for neuroimaging-style analysis using external tooling, though PhysioNet itself does not mandate or organise data around BIDS. The flagship MIMIC databases (de-identified critical-care records from Beth Israel Deaconess Medical Center) are its most-used holding, distributed under credentialed access.
Access model
Datasets carry one of several access tiers: open access (freely downloadable), credentialed access (requiring a verified researcher account and a signed data use agreement, as for MIMIC), and restricted access. The credentialed data use agreement prohibits redistribution to third parties, and as of 2025 certain controlled-access datasets carry additional restrictions under the U.S. Department of Justice Data Security Program.
Connections
- accepts: EDF (EEG and polysomnography signal data)
- accepts: DICOM (MIMIC-CXR chest-radiograph database)
Resources
- https://physionet.org
- https://physionet.org/about/database/ (dataset catalogue)
- https://wfdb.io (WFDB software libraries and format)
- https://doi.org/10.1161/01.CIR.101.23.e215 (Goldberger et al. 2000, founding PhysioNet/PhysioBank/PhysioToolkit paper)