Best Strategy to Define a Multiple Source Model
|Modules||BESA Research Basic or higher|
|Version||5.2 or higher|
The best strategy to define a multiple source model is usually a sequential spatio-temporal approach. However, this may depend on the data to be analyzed. For strategies and examples, please read the key papers on multiple source analysis in Publications/Methods.
The following guidelines apply to discrete source analysis in general:
- The idea behind source analysis is to place or fit sources into the brain at all regions contributing to the data. Localization/fitting is hypothesis testing: The computed source waveforms separate the modeled brain activities and answer the question if and when activity takes place in the modeled brain region. Thus BESA Research reconstructs the brain activity with high temporal resolution.
- Synchronous activity over several square cm can be modeled by a single equivalent source.
- Sources that are close together, have the same orientation and synchronous activity are indistinguishable.
- Sources that are close together but have different orientations and non-synchronous activity can well be separated by a discrete source model.
- Source waveforms are relatively insensitive to variations in dipole location. Therefore a single regional source can accurately model activity of multiple gray matter patches in its vicinity. Regional sources should have a mutual distance of approximately 3cm or more between each other to prevent crosstalk of their source waveforms.
- Regional sources tend to provide more reliable solutions in noisy data as compared to single dipoles, because during a fit of a regional source only its location needs to be determined (no orientation has to be fitted).
- Very often the brain responds to a stimulus with bilateral symmetrical activation. Therefore, a pair of symmetrical regional sources is often a good initial source configuration. The resulting source waveforms will then indicate if the activity is really bilateral.
- The criterion for a good source model is not solely the residual variance. Rather, the following criteria should be met:
- The source model should be in agreement with proven knowledge about the underlying brain activity (e.g. bilateral activation of the auditory cortex after an auditory stimulus as opposed to a single central brain area).
- The source waveforms are an important indicator for the quality of a source model: When the time course of the obtained source waveforms for a (pair of) source(s) are distinctly different from those of other sources, this indicates that the corresponding source correctly models distinct brain activity. On the contrary, when two source waveforms show nearly the same time course, it should be checked if they pick up activity of a brain region that is not part of the model, rather than truly modeling brain activity generated at their source locations.
- PCA can be helpful in determining the minimum number of sources required to adequately model the data.