SCCN Home
EEGLAB Home  

A5.1. Component localization by equivalent dipoles: A simple example

Equivalent dipole source localization by BESA (MEGIS Software GmbH, Munich) can be applied to independent component maps under EEGLAB. The following window shows five independent component scalp maps from five different subjects. Each of these components was labelled by the experimenters as representing a left-occipital alpha rhythm generator.

BESA dipole localization applied to these component maps, accounted for each map with a single equivalent dipole in a three-shell spherical model with low residual variance (R.V.)  Note that unlike in many BESA analyses, we rarely need to ignore signals at some scalp electrodes since ICA typically isolates artifactual signals into separate independent components, and independent components likely account mainly for activity with a connected synchronous cortical domain whose scalp projection should match that of a single equivalent dipole. (Note: This does not mean, however, that some independent components may be generated by tightly linked bilateral activity in domains joined through e.g. corpus callosum. In this case, the source may often be well modelled by two symmetricly placed equivalent dipoles).

A5.2. How to localize components using BESA

BESA is a commercial software (www.besa.de) developed by Michael Scherg and colleagues for localizing brain source equivalent dipoles using a three-shell spherical model.

We have developed Matlab functions to automatically export/import and display BESA dipole information. To localize components, you must run BESA 2.x, 3.x, 99 or BESA 2000 Since BESA 99 and 2000 do not support macro files, if you use these versions you must also download BESA 3.0 (the previous version running macros) from the BESA web site (see below) .

BESA 3.0 installation for BESA 2000 and BESA 99 users

1 - Download BESA30 DOS-Window Upgrade from 2.x  from the BESA web site (even if you don't have BESA 2.x)

2 - Install the hardlock driver for BESA 2000 (on the BESA 2000 CD. Note: you must usually go through the whole installation of BESA 2000 to install the hardlock driver).

3 - Download the BESA30 for BESA2000-Hardlock file and place it in the BESA 3.0 directory (if you only have a BESA 99 license, you should download BESA30 for BESA99-EEG-Hardlock or BESA30 for BESA99-MEG+EEG-Hardlock).

4 - Click on the file downloaded in (3) to start BESA3.0

5 - The BESA3.0 manual is not provided on the Internet. We use the BESA 2.2 Manual and find that most of the commands also work in version 3.0.

Install EEGLAB plugin

Then download the BESA plugin for EEGLAB (go to the plugin page, download the BESA plugin, and uncompress it in the EEGLAB plugin folder). EEGLAB BESA functions can also be used with BESA 2.x. However, BESA 2.x cannot import more than 66 channels, and has some restrictions about path lengths under Windows, we strongly recommend the use of BESA 3.0.

A5.3. Exporting component information

You must first restart EEGLAB with the BESA-function menu included by typing

>>  eeglab besa

Note that this call will not erase your current dataset(s), but will simply add a new item to the Tools menu: Tools > Localize dipoles using BESA.

After ICA components have been computed, select menu item Tools > Localize dipoles using BESA > Export components to BESA. This calls the function besaexport(). The following window will appear:

In the first edit box, enter the indices of the components to be localized using BESA. The other edit boxes are self explanatory. In case you choose to export the electrode locations from EEGLAB (by leaving blank the 4th edit box), it is very important to check that the polar coordinates have not been shrunk by any other program (e.g., the Fz radius value should be 0.25). If you imported a file with 3-D coordinates into EEGLAB, that should not be an issue. Note that under Windows the last edit box will not appear. After pressing OK, the following text will be displayed on the command line:

Accessing directory...
Exporting electrode locations ...
Exporting component maps...
Exporting parameter file...
Writing macro file...
NOW RUN BESA, PRESS "A" (for automatic processing) and enter filename "E:\besatmp\macro"

Now, run BESA on your Windows machine. In BESA, type "A" and then enter the filename given above ("E:\besatmp\macro"). Dipole localization will be performed automatically. Since the number of characters in each BESA macro file is limited, the EEGLAB Matlab function generates many macro files that are called by parent macro files. In this way, up to 900 components can be localized at a time. Also note that minimizing the BESA window will greatly speed up the computation.

Occasionally BESA will not optimize the dipole location (e.g., when residual variance remains 98% or more). If this happens, simply rerun the macro corresponding to the badly localized component (for instance macro10.aut for component 10).

A5.4. Importing BESA component locations

Select menu item Tools > Localize dipoles using BESA > Import dipoles from BESA (calling the function besaimport()). The following interactive window will pop up.

 

In the upper edit box, you must provide the location of the directory in which the dipole information has been saved by BESA. In the lower edit box, you can either (1) import dipoles and erase all previous information ("no" option), (2) append the newly computed information ("yes" option), or (3) only import dipoles for components that do not have yet associated dipoles ("componly" option). Press OK and the function will return information about the components that have been imported:

Reading component 1 (1 dipoles) residual variance 2.09 %
Reading component 2 (1 dipoles) residual variance 15.67 %
Reading component 3 (1 dipoles) residual variance 1.94 %
Reading component 4 (1 dipoles) residual variance 1.23 %
Reading component 5 (1 dipoles) residual variance 3.49 %
Reading component 6 (1 dipoles) residual variance 12.88 %

Note: We usually only trust dipole locations with low residual variance (e.g., below 5%, though this figure may vary with number of channels, montage, etc.).

A5.5. Visualizing dipole locations

There are four menu items that visualize equivalent dipole locations. They all call the same function dipplot() with different options. Here, we illustrate two of these four menu items: First, Tools > Localize dipoles using BESA > Plot dipoles on BESA head.

Here, the dipoles are plotted in 3-D (Note: Use the buttons on the bottom-left to select a specific view). Note that the background images were captured from BESA and the scaling was adjusted manually using visual fitting of a sphere mesh and of the BESA images (press the "Mesh on" button (lower left) to display the mesh sphere). Note that the barlength of the dipole varies with the absolute component amplitude.

We also fit the head sphere model to an average-brain MRI average images (available for download from the ICBM project). MRI images were first shrunk along the X or Y axis to match the 3-D sphere mesh. We then changed the aspect ratio of the axis so that the MRI images' original aspect ratio was preserved (and the sphere mesh became an ellipsoid). A dipole location summary, plotted on the top of the MRI image, can be shown using menu item Tools > Localize dipoles using BESA > Plot dipoles on BESA head.

Finally, by calling the plotting function from the command line

>>  dipplot(EEG.sources);

many more options can be specified and one or more equivalent dipoles can be plotted individually for each component.

A5.6. Miscellany

Using other dipole localization software

You may use the besaexport() function to export the electrode localization file and the component activations, then try to localize dipoles using some other localization software. If you succeed in doing this, or if you write dedicated functions for this, please contact us.

Localizing dipoles from ERP averages or continuous EEG

Because of the theoretical and practical advantages of using ICA for source localization, we have only implemented equivalent dipole localization for independent components. However, localizing ERP or EEG time series would simply involve exporting the data to BESA. For instance, to export an ERP raw data file that BESA can read, type in

>>  ERP = mean(EEG.data,3);
>>  save -ascii ERP.raw ERP

You may also use the electrode location file generated by besaexport().

Dipole localization using DIPFIT

Tutorial index