AnyWave:ICA

From WikiMEG
Jump to: navigation, search

ICA Plug-in

This plugin will extract Independent Components from signals.
The algorithm used is the Infomax algorithm.
This plug-in will run in the background, doing the computation and then a file will be created containing the resulting matrices: (mixing and unmixing matrices).
The only user interface is the one which let the user set all the settings before the computation is done.

ICA ui.png

Algorithm

The plug-in was designed to implement several algorithms but only infomax is implemented so far.

Filtering

The user can specify a frequency bandwidth to use by defining a low pass and a high pass filtering.
Note that decimation will be applied depending on the low pass filtering, to shorten the calculation time and minimize the memory usage.

Data Input

The user must choose on what type of signal the computation will be done (EEG or MEG).
By default, bad channels are excluded. Finally, markers can be used to remove bad part of the signals (artifacts, etc.), the user must first add markers by itself or using another plug-in with an explicit name.

Data Output

The number of components is by default limited to 50 but all the components can be extracted at the cost of a longer calculation.
By default, a result file will be created in the data folder. This file can be imported afterward in AnyWave by using the Load ICA Components feature.

Batch mode

This plugin can run in command line or batch mode.
Here are the json keys the plugin will use:

{
"plugin": "ica", 
"comp" : 50,
"downsampling" : false,
"modality" : "MEG",
"use_markers" : [ "selection1", "selection2"],
"skip_markers" : [ "artefact", "artefacted" ]
}
key description requirement default behavior
plugin Name of plugin MANDATORY n/a
modality channel modality (MEG, EEG, ...) MANDATORY n/a
downsampling enable/disable downsampling of data OPTIONAL default is true
comp number of expected components (PCA applied if comp < n channels) OPTIONAL if not specified, comp will be the number of channels of the modality.
skip_bad true if ignoring bad channels OPTIONAL default is true


example:

anywave --run ica.json --hp 1 --lp 70 --input_file /path_to_file/myfile.dat

Note: some keys can also be specified as command line options (anywave --hp 1 --lp 45 --input_file /path/to/data.meg)

  • lp, hp : defines the filters to apply to the data.
  • input_file: path to the data file.
  • use_markers : list of markers to use as data selections.
  • skip_markers : list of markers to avoid (artefacted parts, for example).
  • output_dir: folder where to store the results.