Datasets Overview

Note

Contributing datasets to MNE-Python

Do not hesitate to contact MNE-Python developers on the MNE Forum to discuss the possibility of adding more publicly available datasets.

All the dataset fetchers are available in mne.datasets. To download any of the datasets, use the data_path (fetches full dataset) or the load_data (fetches dataset partially) functions.

All fetchers will check the default download location first to see if the dataset is already on your computer, and only download it if necessary. The default download location is also configurable; see the documentation of any of the data_path functions for more information.

Sample

mne.datasets.sample.data_path()

These data were acquired with the Neuromag Vectorview system at MGH/HMS/MIT Athinoula A. Martinos Center Biomedical Imaging. EEG data from a 60-channel electrode cap was acquired simultaneously with the MEG. The original MRI data set was acquired with a Siemens 1.5 T Sonata scanner using an MPRAGE sequence.

Note

These data are provided solely for the purpose of getting familiar with the MNE software. The data should not be used to evaluate the performance of the MEG or MRI system employed.

In this experiment, checkerboard patterns were presented to the subject into the left and right visual field, interspersed by tones to the left or right ear. The interval between the stimuli was 750 ms. Occasionally a smiley face was presented at the center of the visual field. The subject was asked to press a key with the right index finger as soon as possible after the appearance of the face.

Trigger codes for the sample data set.

Name		Contents
LA	1	Response to left-ear auditory stimulus
RA	2	Response to right-ear auditory stimulus
LV	3	Response to left visual field stimulus
RV	4	Response to right visual field stimulus
smiley	5	Response to the smiley face
button	32	Response triggered by the button press

Contents of the data set

The sample data set contains two main directories: MEG/sample (the MEG/EEG data) and subjects/sample (the MRI reconstructions). In addition to subject sample, the MRI surface reconstructions from another subject, morph, are provided to demonstrate morphing capabilities.

Contents of the MEG/sample directory.

File	Contents
sample/audvis_raw.fif	The raw MEG/EEG data
audvis.ave	A template script for off-line averaging
auvis.cov	A template script for the computation of a noise-covariance matrix

Overview of the contents of the subjects/sample directory.

File / directory	Contents
bem	Directory for the forward modelling data
bem/watershed	BEM surface segmentation data computed with the watershed algorithm
bem/inner_skull.surf	Inner skull surface for BEM
bem/outer_skull.surf	Outer skull surface for BEM
bem/outer_skin.surf	Skin surface for BEM
sample-head.fif	Skin surface in fif format for mne_analyze visualizations
surf	Surface reconstructions
mri/T1	The T1-weighted MRI data employed in visualizations

The following preprocessing steps have been already accomplished in the sample data set:

• The MRI surface reconstructions have been computed using the FreeSurfer software.

• The BEM surfaces have been created with the watershed algorithm, see Using the watershed algorithm.

The sample dataset is distributed with fsaverage for convenience.

UCL OPM Auditory

mne.datasets.ucl_opm_auditory.data_path().

A basic auditory evoked field experiment using an OPM setup from FIL at UCL. See [1] for details.

Examples

• Preprocessing optically pumped magnetometer (OPM) MEG data

Brainstorm

Dataset fetchers for three Brainstorm tutorials are available. Users must agree to the license terms of these datasets before downloading them. These files are recorded in a CTF 275 system and are provided in native CTF format (.ds files).

Auditory

mne.datasets.brainstorm.bst_raw.data_path().

Details about the data can be found at the Brainstorm auditory dataset tutorial.

Examples

• Working with CTF data: the Brainstorm auditory dataset: Partially replicates the original Brainstorm tutorial.

Resting state

mne.datasets.brainstorm.bst_resting.data_path()

Details can be found at the Brainstorm resting state dataset tutorial.

Examples

• Compute envelope correlations in source space

Median nerve

mne.datasets.brainstorm.bst_raw.data_path()

Details can be found at the Brainstorm median nerve dataset tutorial.

Examples

• Brainstorm raw (median nerve) dataset

SPM faces

mne.datasets.spm_face.data_path()

The SPM faces dataset contains EEG, MEG and fMRI recordings on face perception.

Examples

• From raw data to dSPM on SPM Faces dataset Full pipeline including artifact removal, epochs averaging, forward model computation and source reconstruction using dSPM on the contrast: “faces - scrambled”.

EEGBCI motor imagery

mne.datasets.eegbci.load_data()

The EEGBCI dataset is documented in [2]. The data set is available at PhysioNet [3]. The dataset contains 64-channel EEG recordings from 109 subjects and 14 runs on each subject in EDF+ format. The recordings were made using the BCI2000 system. To load a subject, do:

from mne.io import concatenate_raws, read_raw_edf
from mne.datasets import eegbci
raw_fnames = eegbci.load_data(subject, runs)
raws = [read_raw_edf(f, preload=True) for f in raw_fnames]
raw = concatenate_raws(raws)

Examples

• Motor imagery decoding from EEG data using the Common Spatial Pattern (CSP)

Somatosensory

mne.datasets.somato.data_path()

This dataset contains somatosensory data with event-related synchronizations (ERS) and desynchronizations (ERD).

Examples

• Frequency and time-frequency sensor analysis

• Compute source power using DICS beamformer

• Explore event-related dynamics for specific frequency bands

Multimodal

mne.datasets.multimodal.data_path()

This dataset contains a single subject recorded at Otaniemi (Aalto University) with auditory, visual, and somatosensory stimuli.

Examples

• Getting averaging info from .fif files

fNIRS motor

mne.datasets.fnirs_motor.data_path()

This dataset contains a single subject recorded at Macquarie University. It has optodes placed over the motor cortex. There are three conditions:

• tapping the left thumb to fingers

• tapping the right thumb to fingers

• a control where nothing happens

The tapping lasts 5 seconds, and there are 30 trials of each condition.

Examples

• Preprocessing functional near-infrared spectroscopy (fNIRS) data

High frequency SEF

mne.datasets.hf_sef.data_path()

This dataset contains somatosensory evoked fields (median nerve stimulation) with thousands of epochs. It was recorded with an Elekta TRIUX MEG device at a sampling frequency of 3 kHz. The dataset is suitable for investigating high-frequency somatosensory responses. Data from two subjects are included with MRI images in DICOM format and FreeSurfer reconstructions.

Examples

• high-frequency SEF responses.

Visual 92 object categories

mne.datasets.visual_92_categories.data_path().

This dataset is recorded using a 306-channel Neuromag vectorview system.

Experiment consisted in the visual presentation of 92 images of human, animal and inanimate objects either natural or artificial [4]. Given the high number of conditions this dataset is well adapted to an approach based on Representational Similarity Analysis (RSA).

Examples

• Representational Similarity Analysis (RSA): Partially replicates the results from [4].

mTRF Dataset

mne.datasets.mtrf.data_path().

This dataset contains 128 channel EEG as well as natural speech stimulus features, which is also available here.

The experiment consisted of subjects listening to natural speech. The dataset contains several feature representations of the speech stimulus, suitable for using to fit continuous regression models of neural activity. More details and a description of the package can be found in [5].

Examples

• Receptive Field Estimation and Prediction: Partially replicates the results from [5].

Kiloword dataset

mne.datasets.kiloword.data_path().

This dataset consists of averaged EEG data from 75 subjects performing a lexical decision task on 960 English words [6]. The words are richly annotated, and can be used for e.g. multiple regression estimation of EEG correlates of printed word processing.

KIT phantom dataset

mne.datasets.phantom_kit.data_path().

This dataset was obtained with a phantom on a KIT system at Macquarie University in Sydney, Australia.

Examples

• KIT phantom dataset tutorial

4D Neuroimaging / BTi dataset

mne.datasets.phantom_4dbti.data_path().

This dataset was obtained with a phantom on a 4D Neuroimaging / BTi system at the MEG center in La Timone hospital in Marseille.

Examples

• 4D Neuroimaging/BTi phantom dataset tutorial

Kernel OPM phantom dataset

mne.datasets.phantom_kernel.data_path().

This dataset was obtained with a Neuromag phantom in a Kernel Flux (720-sensor) system at ILABS at the University of Washington. Only 7 out of 42 possible modules were active for testing purposes, yielding 121 channels of data with limited coverage (mostly occipital and parietal).

Examples

• Kernel OPM phantom data

OPM

mne.datasets.opm.data_path()

OPM data acquired using an Elekta DACQ, simply piping the data into Elekta magnetometer channels. The FIF files thus appear to come from a TRIUX system that is only acquiring a small number of magnetometer channels instead of the whole array.

The OPM coil_type is custom, requiring a custom coil_def.dat. The new coil_type is 9999.

OPM co-registration differs a bit from the typical SQUID-MEG workflow. No -trans.fif file is needed for the OPMs, the FIF files include proper sensor locations in MRI coordinates and no digitization of RPA/LPA/Nasion. Thus the MEG<->Head coordinate transform is taken to be an identity matrix (i.e., everything is in MRI coordinates), even though this mis-identifies the head coordinate frame (which is defined by the relationship of the LPA, RPA, and Nasion).

Triggers include:

• Median nerve stimulation: trigger value 257.

• Magnetic trigger (in OPM measurement only): trigger value 260. 1 second before the median nerve stimulation, a magnetic trigger is piped into the MSR. This was to be able to check the synchronization between OPMs retrospectively, as each sensor runs on an independent clock. Synchronization turned out to be satisfactory.

Examples

• Optically pumped magnetometer (OPM) data

• Compute source power spectral density (PSD) of VectorView and OPM data

The Sleep PolySomnoGraphic Database

mne.datasets.sleep_physionet.age.fetch_data() mne.datasets.sleep_physionet.temazepam.fetch_data()

The sleep PhysioNet database contains 197 whole-night PolySomnoGraphic sleep recordings, containing EEG, EOG, chin EMG, and event markers. Some records also contain respiration and body temperature. Corresponding hypnograms (sleep patterns) were manually scored by well-trained technicians according to the Rechtschaffen and Kales manual, and are also available. If you use these data please cite [7] and [3].

Examples

• Sleep stage classification from polysomnography (PSG) data

Reference channel noise MEG data set

mne.datasets.refmeg_noise.data_path().

This dataset was obtained with a 4D Neuroimaging / BTi system at the University Clinic - Erlangen, Germany. There are powerful bursts of external magnetic noise throughout the recording, which make it a good example for automatic noise removal techniques.

Examples

• Find MEG reference channel artifacts

Miscellaneous Datasets

These datasets are used for specific purposes in the documentation and in general are not useful for separate analyses.

fsaverage

mne.datasets.fetch_fsaverage()

For convenience, we provide a function to separately download and extract the (or update an existing) fsaverage subject.

Examples

EEG forward operator with a template MRI

Infant template MRIs

mne.datasets.fetch_infant_template()

This function will download an infant template MRI from [8] along with MNE-specific files.

ECoG Dataset

mne.datasets.misc.data_path(). Data exists at /ecog/.

This dataset contains a sample electrocorticography (ECoG) dataset. It includes two grids of electrodes and ten shaft electrodes with simulated motor data (actual data pending availability).

Examples

• How to convert 3D electrode positions to a 2D image: Demonstrates how to project a 3D electrode location onto a 2D image, a common procedure in ECoG analyses.

• Locating intracranial electrode contacts: Demonstrates how to use a graphical user interface to locate electrode contacts as well as warp them to a common atlas.

sEEG Dataset

mne.datasets.misc.data_path(). Data exists at /seeg/.

This dataset contains a sample stereoelectroencephalography (sEEG) dataset. It includes 21 shaft electrodes during a two-choice movement task on a keyboard.

Examples

• Locating intracranial electrode contacts: Demonstrates how to use a graphical user interface to locate electrode contacts as well as warp them to a common atlas.

• Working with sEEG data: Demonstrates ways to plot sEEG anatomy and results.

LIMO Dataset

mne.datasets.limo.load_data().

In the original LIMO experiment (see [9]), participants performed a two-alternative forced choice task, discriminating between two face stimuli. Subjects discriminated the same two faces during the whole experiment. The critical manipulation consisted of the level of noise added to the face-stimuli during the task, making the faces more or less discernible to the observer.

The presented faces varied across a noise-signal (or phase-coherence) continuum spanning from 0 to 100% in increasing steps of 10%. In other words, faces with high phase-coherence (e.g., 90%) were easy to identify, while faces with low phase-coherence (e.g., 10%) were hard to identify and by extension hard to discriminate.

Examples

• Single trial linear regression analysis with the LIMO dataset: Explores data from a single subject of the LIMO dataset and demonstrates how to fit a single trial linear regression using the information contained in the metadata of the individual datasets.

ERP CORE Dataset

mne.datasets.erp_core.data_path()

The original ERP CORE dataset [10] contains data from 40 participants who completed 6 EEG experiments, carefully crafted to evoke 7 well-known event-related potential (ERP) components.

Currently, the MNE-Python ERP CORE dataset only provides data from one participant (subject 001) of the Flankers paradigm, which elicits the lateralized readiness potential (LRP) and error-related negativity (ERN). The data provided is not the original data from the ERP CORE dataset, but rather a slightly modified version, designed to demonstrate the Epochs metadata functionality. For example, we already set the references and montage correctly, and stored events as Annotations. Data is provided in FIFF format.

Examples

• Auto-generating Epochs metadata: Learn how to auto-generate Epochs metadata, and visualize the error-related negativity (ERN) ERP component.

SSVEP

mne.datasets.ssvep.data_path()

This is a simple example dataset with frequency tagged visual stimulation: N=2 participants observed checkerboards patterns inverting with a constant frequency of either 12.0 Hz of 15.0 Hz. 10 trials of 20.0 s length each. 32 channels wet EEG was recorded.

Data format: BrainVision .eeg/.vhdr/.vmrk files organized according to BIDS standard.

Examples

• Frequency-tagging: Basic analysis of an SSVEP/vSSR dataset

EYELINK

mne.datasets.eyelink.data_path()

Two small example datasets of eye-tracking data from SR Research EyeLink.

EEG-Eyetracking

mne.datasets.eyelink.data_path(). Data exists at /eeg-et/.

Contains both EEG (EGI) and eye-tracking (ASCII format) data recorded from a pupillary light reflex experiment, stored in separate files. 1 participant fixated on the screen while short light flashes appeared. Event onsets were recorded by a photodiode attached to the screen and were sent to both the EEG and eye-tracking systems.

Examples

• Working with eye tracker data in MNE-Python

Freeviewing

mne.datasets.eyelink.data_path(). Data exists at /freeviewing/.

Contains eye-tracking data (ASCII format) from 1 participant who was free-viewing a video of a natural scene. In some videos, the natural scene was pixelated such that the people in the scene were unrecognizable.

Examples

• Plotting eye-tracking heatmaps in MNE-Python

References

[1]

Robert A. Seymour, Nicholas Alexander, Stephanie Mellor, George C. O’Neill, Tim M. Tierney, Gareth R. Barnes, and Eleanor A. Maguire. Interference suppression techniques for OPM-based MEG: Opportunities and challenges. NeuroImage, 247:118834, February 2022. doi:10.1016/j.neuroimage.2021.118834.

[2]

Gerwin Schalk, Dennis J. McFarland, Thilo Hinterberger, Niels Birbaumer, and Jonathan R. Wolpaw. BCI2000: a general-purpose brain-computer interface (BCI) system. IEEE Transactions on Biomedical Engineering, 51(6):1034–1043, 2004. doi:10.1109/TBME.2004.827072.

[3] (1,2)

Ary L. Goldberger, Luis A. N. Amaral, Leon Glass, Jeffrey M. Hausdorff, Plamen Ch. Ivanov, Roger G. Mark, Joseph E. Mietus, George B. Moody, Chung-Kang Peng, and H. Eugene Stanley. PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation, 2000. doi:10.1161/01.CIR.101.23.e215.

[4] (1,2)

Radoslaw Martin Cichy, Dimitrios Pantazis, and Aude Oliva. Resolving human object recognition in space and time. Nature Neuroscience, 17(3):455–462, 2014. doi:10.1038/nn.3635.

[5] (1,2)

Michael J. Crosse, Giovanni M. Di Liberto, Adam Bednar, and Edmund C. Lalor. The multivariate temporal response function (mTRF) toolbox: a MATLAB toolbox for relating neural signals to continuous stimuli. Frontiers in Human Neuroscience, 2016. doi:10.3389/fnhum.2016.00604.

[6]

Stéphane Dufau, Jonathan Grainger, Katherine J. Midgley, and Phillip J. Holcomb. A thousand words are worth a picture: snapshots of printed-word processing in an event-related potential megastudy. Psychological Science, 26(12):1887–1897, 2015. doi:10.1177/0956797615603934.

[7]

B. Kemp, A. H. Zwinderman, B. Tuk, H. A. C. Kamphuisen, and J. J. L. Oberyé. Analysis of a sleep-dependent neuronal feedback loop: the slow-wave microcontinuity of the EEG. IEEE Transactions on Biomedical Engineering, 47(9):1185–1194, 2000. doi:10.1109/10.867928.

[8]

Christian O’Reilly, Eric Larson, John E. Richards, and Mayada Elsabbagh. Structural templates for imaging EEG cortical sources in infants. NeuroImage, 227:117682, 2021. doi:10.1016/j.neuroimage.2020.117682.

[9]

Guillaume A. Rousselet, Carl M. Gaspar, Cyril R. Pernet, Jesse S. Husk, Patrick J. Bennett, and Allison B. Sekuler. Healthy aging delays scalp EEG sensitivity to noise in a face discrimination task. Frontiers in Psychology, 1(19):1–14, 2010. doi:10.3389/fpsyg.2010.00019.

[10]

Emily S. Kappenman, Jaclyn L. Farrens, Wendy Zhang, Andrew X. Stewart, and Steven J. Luck. ERP CORE: an open resource for human event-related potential research. NeuroImage, 225:117465, 2021. doi:10.1016/j.neuroimage.2020.117465.