The sample data set


This Chapter gives a detailed description of the processing of a sample data set, which can be employed to familiarize with the workflow described in The typical M/EEG workflow.


Going through the analysis exercise in this chapter is not a substitute for reading other chapters of this manual and understanding the concepts underlying MNE software.


The MNE software is accompanied by a sample data set which includes the MRI reconstructions created with FreeSurfer and the an MEG/EEG data set. 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.


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

In the MEG/EEG experiment, checkerboard patterns were presented 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. A listing of the corresponding trigger codes is provided in Trigger codes for the sample data set.

Trigger codes for the sample data set.





Response to left-ear auditory stimulus



Response to right-ear auditory stimulus



Response to left visual field stimulus



Response to right visual field stimulus



Response to the smiley face



Response triggered by the button press

Setting up

The sample dataset can be downloaded automatically by doing:

>>> mne.datasets.sample.data_path(verbose=True)  

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). An overview of the data provided is given in Tables Contents of the MEG/sample directory. and Overview of the contents of the subjects/sample directory.. In addition to subject sample , the MRI surface reconstructions from another subject, morph , are provided to demonstrate the morphing capabilities of the MNE software.

Contents of the MEG/sample directory.




The raw MEG/EEG data


A template script for off-line averaging


A template script for the computation of a noise-covariance matrix

Overview of the contents of the subjects/sample directory.

File / directory



Directory for the forward modelling data


BEM surface segmentation data computed with the watershed algorithm


Inner skull surface for BEM


Outer skull surface for BEM


Skin surface for BEM


Skin surface in fif format for mne_analyze visualizations


Surface reconstructions


The T1-weighted MRI data employed in visualizations

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

Setting up subject-specific data

Structural MRIs

To set up the structural MRIs for processing with the Neuromag MRI viewer, MRIlab, say


This command sets up the directories subjects/sample/mri/T1-neuromag and subjects/sample/mri/brain-neuromag . For more information, see Setting up anatomical MR images for MRIlab.

Source space

The source space with a 5-mm grid spacing is set up by saying:

mne_setup_source_space --ico -6

This command sets up the source-space related files in directory subjects/sample/bem as described in Setting up the source space.

Boundary-element models

The geometry calculations for the single-layer boundary-element model are accomplished with the command:

mne_setup_forward_model --homog --surf --ico 4

This command sets up the homogeneous BEM-model related files in directory subjects/sample/bem as described in Setting up the boundary-element model. In addition to the homogeneous BEM, you also need the three-layer BEM model, which can be used for both EEG and MEG:

mne_setup_forward_model --surf --ico 4

The above commands employ the , , and triangulation files located in subjects/sample/bem . The option --ico 4 will create a model with 5120 triangles on each surface. Depending on the speed of your computer, the three-layer model may take quite a while to set up.

Setting up a custom EEG layout

A data specific EEG layout will facilitate viewing of the EEG data. The MNE programs mne_browse_raw and mne_analyze look for user-specific layouts in $HOME/.mne/lout . Thus, you can create an EEG layout for the sample data set with the following commands:

mkdir -p $HOME/.mne/lout

cd $SAMPLE/MEG/sample

mne_make_eeg_layout --fif sample_audvis_raw.fif --lout $HOME/.mne/lout/sample-EEG.lout

Please refer to mne_make_eeg_layout for more information on mne_make_eeg_layout .


It is usually sufficient to create one EEG layout for each electrode cap you are using in your experiment rather than using a different layout file for each data file generated using the same cap.

Previewing the data

Previewing your raw data before proceeding to averaging and computation of the current estimates is most important to avoid unintentional errors caused by noisy or dysfunctional channels, frequent eye blinks, inappropriate bandpass filtering etc.

One possible routemap for the preview session is outlined below:

  • Go to the MEG/EEG data directory: cd $SAMPLE/MEG/sample .

  • Launch mne_browse_raw .

  • Open the raw data file sample_audvis_raw.fif from File/Open… .

  • Switch all SSP vectors off from Adjust/Projection… .

  • Set the lowpass filter corner to a high value, e.g., 150 Hz from Adjust/Filter… .

  • Browse through all channels using the selections available under Adjust/Selection… and look for very noisy or flat channels. You should be able to readily identify two such channels among all MEG and EEG channels. You may need to click Remove DC to reliably associate the noisy or flat waveform with the channel name on the left. Also, experiment with switching the EEG average reference projection on and off and you will notice that the EEG bad channel cannot be seen after the projection.

  • Mark the channels you identified in step 6. bad for this viewing session by clicking on their channel names on the left. You can save the bad channel selection to the file from File/Apply bad channels . Bad channel marking can be removed by clicking on their channel names again and selecting File/Apply bad channels . Alternatively, you can use the utility mne_mark_bad_channels to set a bad channel selection, see mne_mark_bad_channels.

  • Switch the projections back on and change filter to a 40-Hz lowpass.

  • Compute a preliminary average for the left-ear auditory stimulus:

    • Open the averaging preferences dialog (Adjust/Averaging preferences… .

    • Set the time scale to -100…300 ms.

    • Click on the text next to Average: in the main window and press return. After a while, a topographical display appears with the averaged responses. Notice that the bad channels are not displayed.

    • Change to different layouts from Adjust/Full view layout… . Inspect both the MEG and EEG waveforms.

  • Compute a preliminary average for the right-ear auditory stimulus:

    • Open the averaging preferences.

    • Click on the Trace color… button and change the trace color to something different from the default yellow.

    • Change the text next to Average: to 2 and press return. Average to the right-ear tones will be computed. Compare the to sets of averages and verify that all channels show reasonable data.

  • Go to Windows/Manage averages… and delete the preliminary averages just computed.

After these steps, you are ready to proceed to the actual analysis.

Off-line averaging

Go to directory $SAMPLE/MEG/sample . With help of Description files for off-line averaging, familiarize yourself with the averaging script audvis.ave .

Using the averaging script interactively

You can invoke an averaging script in mne_browse_raw from Process/Average… . Select the audvis.ave script from the file selection box that appears. Once averaging is complete, you can inspect the details of the averaged responses in the Averages window, which appears automatically. You can redisplay it from Windows/Show averages… . The window, which appears when you select Adjust/Manage averages… allows you to:

  • Select which conditions (categories) are displayed.

  • Change the trace colors.

  • Inspect the averaging log.

  • Save the averaged data.

  • Delete this set of averages.


If you decide to save the averages in the interactive mode, use the name sample_audvis-ave.fif for the result.

Using the averaging script in batch mode

The batch-mode version of mne_browse_raw , mne_process_raw can be used for averaging as well. Batch mode averaging can be done with the command:

mne_process_raw --raw sample_audvis_raw.fif `` ``--lowpass 40 --projoff `` ``--saveavetag -ave --ave audvis.ave

See mne_process_raw for command-line options.

As a result of running the averaging script a file called sample_audvis-ave.fif is created. It contains averages to the left and right ear auditory as well as to the left and right visual field stimuli.

Viewing the off-line average

The average file computed in the previous section can be viewed in mne_browse_raw .

To view the averaged signals, invoke mne_browse_raw :

cd $SAMPLE/MEG/sample

mne_browse_raw &

This Section gives only very basic information about the use of mne_browse_raw for viewing evoked-response data. Please consult Browsing raw data with mne_browse_raw for more comprehensive information.

Loading the averages

mne_browse_raw loads all the available data from an average file at once:

  • Select Open evoked… from the File menu.

  • Select the average file sample_audvis-ave.fif file from the list and click OK .

  • A topographical display of the waveforms with gradiometer channels included appears.

Inspecting the auditory data

Select the left and right ear auditory stimulus responses for display:

  • Select Manage averages… from the Adjust menu.

  • Click off all other conditions except the auditory ones.

Set the time scale and baseline:

  • Select Scales… from the Adjust menu.

  • Switch off Autoscale time range and set the Average time range from -200 to 500 ms.

  • Switch on Use average display baseline and set Average display baseline from -200 to 0 ms.

  • Click OK .

You can display a subset of responses from the topographical display by holding the shift key down and dragging with the mouse, left button down. When you drag on the response with just the left button down, the signal timing, and channel name are displayed at the bottom. If the left mouse button is down and you press shift down the time is give both in absolute units and relative to the point where shift was pressed down.

Observe the following:

  • The main deflection occurs around 100 ms over the left and right temporal areas.

  • The left-ear response (shown in yellow) is stronger on the right than on the left. The opposite is true for the right-ear response, shown in red.

Inspecting the visual data

Go back to the Manage averages… dialog and switch all other conditions except the visual ones.

Observe the following:

  • The left and right visual field responses are quite different in spatial distribution in the occipital area.

  • There is a later response in the right parietal area, almost identical to both visual stimuli.


If you have the Neuromag software available, the averaged data can be also viewed in the Neuromag data plotter (xplotter ). See Using Neuromag software for instructions on how to use the Neuromag software at the MGH Martinos Center.

Computing the noise-covariance matrix

Another piece of information derived from the raw data file is the estimate for the noise-covariance matrix, which can be computed with the command:

mne_process_raw --raw sample_audvis_raw.fif --lowpass 40 --projon --savecovtag -cov --cov audvis.cov

Using the definitions in audvis.cov , this command will create the noise-covariance matrix file sample_audvis-cov.fif . In this case the projections are set on. The projection information is then attached to the noise-covariance matrix and will be automatically loaded when the inverse-operator decomposition is computed.


You can study the contents of the covariance matrix computation description file audvis.cov with the help of Description files for covariance matrix estimation.

MEG-MRI coordinate system alignment

The mne_analyze module of the MNE is one option for the coordinate alignment. It uses a triangulated scalp surface to facilitate the alignment.

Initial alignment

Follow these steps to make an initial approximation for the coordinate alignment.

  • Go to directory MEG/sample .

  • Launch mne_analyze

  • Select File/Load digitizer data… and load the digitizer data from sample_audvis_raw.fif .

  • Load an inflated surface for subject sample from File/Load surface…

  • Bring up the viewer window from View/Show viewer…

  • Click Options… in the viewer window. Make the following selections:

    • Switch left and right cortical surface display off.

    • Make the scalp transparent.

    • Switch Digitizer data on.

  • After a while, the digitizer points will be shown. The color of the circles indicates whether the point is inside (blue) or outside (red) of the scalp. The HPI coils are shown in green and the landmark locations in light blue or light red color. The initial alignment is way off!

  • Switch the Digitizer data off to get the big circles out of the way.

  • Bring up the coordinate alignment window from Adjust/Coordinate alignment…

  • Click on the RAP (Right Auricular Point) button. It turns red, indicating that you should select the point from the viewer window. Click at the approximate location of this point in the viewer. The button jumps up, turns to normal color, and the MRI coordinates of the point appear in the text fields next to the button.

  • Proceed similarly for the other two landmark points: Nasion and LAP (Left Auricular Point).

  • Press Align using fiducials . Notice that the coordinate transformation changes from a unit transformation (no rotation, no origin translation) to a one determined by the identified landmark locations. The rotation matrix (upper left 3 x 3 part of the transformation) should have positive values close to one on the diagonal. Three is a significant rotation around the x axis as indicated by elements (3,2) and (2,3) of the rotation matrix. The x and y values of the translation should be small and the z value should be negative, around -50 mm. An example of an initial coordinate transformation is shown in Example of an initial coordinate alignment..

  • Make the Digitizer data again visible from the options of the viewer window. Note that the points are now much closer to the scalp surface.

Example of an initial coordinate alignment

Example of an initial coordinate alignment.

Refining the coordinate transformation

Before proceeding to the refinement procedure, it is useful to remove outlier digitizer points. When you rotate the image in the viewer window, you will notice that there is at least one such point over the right cheek. To discard this point:

  • Click on Discard in the Adjust coordinate alignment window.

  • Enter 10 for the distance of the points to be discarded.

  • Click done. The outlier point disappears.

The coordinate transformation can be adjusted manually with the arrow buttons in the middle part of the Adjust coordinate alignment dialog. These buttons move the digitizer points in the directions indicated by the amount listed next to each of the buttons.

An automatic iterative procedure, Iterative Closest Point (ICP) matching is also provided. At each iteration step

  • For each digitizer point, transformed from MEG to the MRI coordinate frame, the closest point on the triangulated surface is determined.

  • The best coordinate transformation aligning the digitizer points with the closest points on the head surface is computed.

In step 2 of the iteration, the nasion is assigned five times the weight of the other points since it can be assumed that the nasion is the easiest point to identify reliably from the surface image.

The ICP alignment can be invoked by entering the desired number of iterations next to the ICP align button followed by return or simply pressing the ICP align button. The iteration will converge in 10 to 20 steps.


Use the ICP alignment option in mne_analyze with caution. The iteration will not converge to a reasonable solution unless and initial alignment is performed first according to Initial alignment. Outlier points should be excluded as described above. No attempt is made to compensate for the possible distance of the digitized EEG electrode locations from the scalp.

Saving the transformation

To create a MRI fif description file which incorporates the coordinate transformation click Save MRI set in the Adjust coordinate alignment dialog. This will create the MRI set file in the $SUBJECTS_DIR/sample/mri/T1-neuromag/sets directory, which was created by mne_setup_mri_data , see Structural MRIs. The file will be called

COR- <username>- <date>- <time> .fif

where <username> is your login name.

You can also save transformation to a fif file through the Save… button. If the file does not exist, it will only contain the coordinate transformation. If the file exists it will be inserted to the appropriate context. An existing transformation will not be replaced unless Overwrite existing transform is checked in the save dialog.

Once you have saved the coordinate transformation, press Done and quit mne_analyze (File/Quit ).


If you dismiss the alignment dialog before saving the transformation, it will be lost.

The forward solution

To compute the forward solution, say:

cd $SAMPLE/MEG/sample

mne_do_forward_solution --mindist 5 --spacing oct-6 --bem sample-5120-5120-5120 --meas sample_audvis-ave.fif

This produces an EEG and MEG forward solution with source space points closer than 5 mm to the inner skull surface omitted. The source space created in Source space will be employed. As the output from this command will indicate The forward solution will be stored in file sample_audvis-ave-oct-6-fwd.fif .

This command uses the three-layer BEM model sample-5120-5120-5120-bem-sol.fif created in Boundary-element models. If you want to use the single-compartment BEM sample-5120-bem-sol.fif usable for MEG data only say:

cd $SAMPLE/MEG/sample

mne_do_forward_solution --mindist 5 --spacing oct-6 --meas sample_audvis-ave.fif --bem sample-5120 --megonly

The inverse operator decomposition

The inverse operator information, necessary for the computation of the MNEs and dSPMs is accomplished by the command:

mne_do_inverse_operator --fwd sample_audvis-ave-oct-6-fwd.fif --depth --loose 0.2 --meg --eeg

This produces a depth-weighted inverse operator decomposition with ‘loose’ orientation constraint applied. More details on the convenience script mne_do_inverse_operator are provided in Calculating the inverse operator.

The above command employs both EEG and MEG data. To create separate solution for EEG and MEG, run the commands:

mne_do_inverse_operator --fwd sample_audvis-ave-oct-6-fwd.fif --depth --loose 0.2 --meg


mne_do_inverse_operator --fwd sample_audvis-ave-oct-6-fwd.fif --depth --loose 0.2 --eeg


If you were using a single-compartment BEM to compute the forward solution, you can only compute the MEG inverse operator.

Interactive analysis

The most exciting part of this exercise is to explore the data and the current estimates in mne_analyze . This section contains some useful steps to get you started. A lot of information about the capabilities of mne_analyze is given in Interactive analysis with mne_analyze. Batch-mode processing with mne_make_movie is discussed in Producing movies and snapshots. Cross-subject averaging is covered in Morphing and averaging.

Before launching mne_analyze it is advisable to go to the directory MEG/sample . The current working directory can be also changed from mne_analyze .

Getting started

Launch mne_analyze . Select Help/On GLX… , which brings up a window containing Open GL rendering context information. If first line in the information dialog that pops up says Nondirect rendering context instead of Direct rendering context you will experience slow graphics performance. To fix this, your system software, graphics adapter or both need to be updated. Consult a computer support person for further information.

Load surfaces

It is reasonable to start the analysis by loading the display surfaces: choose the inflated surface for subject sample from the dialog that appears when you select File/Load surface… .

Load the data

Select File/Open… . Select sample_audvis-ave.fif as your data file and select the Left auditory data set. Select the inverse operator sample_audvis-ave-oct-6-meg-eeg-inv.fif and press OK . After a while the signals appear in the sample waveform and topographical displays. Click on the N100m peak in the auditory response. A dSPM map appears in the main surface display.

Show field and potential maps

Select Windows/Show viewer… . After a while the viewer window appears. Click on the N100m peak again. Once the field map preparation computations are complete, the magnetic field and potential maps appear. Investigate the viewer window options with help of The viewer.

Show current estimates

The options affecting the current estimates are accessible from Adjust/Estimate parameters… . With help of Working with current estimates, investigate the effects of the parameter settings.

Labels and timecourses

While in directory MEG/sample , create a directory called label :

mkdir label

Using the information in Creating new label files, create two labels A-lh.label and A-rh.label in the approximate location of the left and right auditory cortices. Save these labels in the newly created label directory.

Load all labels from the label directory and investigate the timecourses in these two labels as well as at individual vertices. Information on label processing can be found from Inquiring timecourses.


Goto to $SUBJECTS_DIR and create the directory morph-maps . Load the inflated surface for subject morph as the morphing surfaces. Try switching between the original and morphing surfaces. More information about morphing is available in Morphing and in Morphing and averaging.

There is also a left-hemisphere occipital patch file available for subject morph . Load a righ-hemifield visual response instead of the auditory one and investigate mapping of the current estimates on the patch.