eeg_ica: EEG signal decomposition using Independent Component Analysis...

In bnicenboim/eeguana: Flexible Manipulation of EEG Data

EEG signal decomposition using Independent Component Analysis (ICA)

Description

This function returns an extended eeg_lst, eeg_ica_lst, with the mixing and unmixing matrix of every recording. It is possible to visualize the topography of the components with plot_components(). In order to extract the amplitude of the components with respect to time use eeg_ica_show(), see examples section. To remove the unwanted components, use eeg_ica_keep().

Usage

eeg_ica(
  .data,
  ...,
  .ignore = .type == "artifact",
  .method = fast_ICA,
  .config = list()
)

Arguments

.data	An eeg_lst object
...	Channels to include in ICA transformation. All the channels by default, but eye channels and reference channels should be removed.
.ignore	Events that should be ignored in the ICA, set to NULL for not using the events table.
.method	Methods from different packages: fast_ICA, a wrapper of fastICA::fastICA, (default), and some more experimental .methods: fast_ICA2 and adapt_fast_ICA adapted from fICA package. It can also accept a custom function, see details.
.config	Other parameters passed in a list to the ICA method. See the documentation of the relevant .method.

Details

It is possible to also use a custom function in the method argument. The function should return a list that with A (mixing matrix), consistent with the formulation X = S %*% A, where X is matrix of N_samples by N_channels and/or W (unmixing matrix), consistent with the formulation X %*% W = S. Some packages with other ICA methods or implementations: steadyICA and ica.

Value

An eeg_ica_lst object

See Also

Other ICA functions: eeg_ica_cor_tbl(), eeg_ica_keep(), eeg_ica_show(), eeg_ica_summary_tbl(), eeg_ica_var_tbl(), plot_components()

Other preprocessing functions: eeg_baseline(), eeg_downsample(), eeg_ica_keep(), eeg_rereference(), eeg_segment(), filt

Examples

# For demonstration only, since ICA won't converge
# Suppressing an important warning:
suppressWarnings(data_faces_10_trials %>%
  eeg_ica(-EOGH, -EOGV, -M1, -M2, .method = fast_ICA, .config = list(maxit = 10)))

## The example can only bu run, if python is properly configured (see reticulate package help)
## Here a python ICA function is used:
## Not run: 
library(reticulate)
use_condaenv("anaconda3") # use the appropriate environment
sk <- import("sklearn.decomposition")
py_fica <- function(x) {
  x <- as.matrix(x)
  ica <- sk$FastICA(whiten = TRUE, random_state = 23L)
  X <- scale(x, scale = FALSE) %>%
    as.matrix(x)
  S <- ica$fit_transform(X)
  W <- t(ica$components_)
  list(W = W)
}
data_ica_py <- eeg_ica(data_faces_10_trials, -EOGH, -EOGV, -M1, -M2, .method = py_fica)

## End(Not run)

bnicenboim/eeguana documentation built on March 16, 2024, 7:21 a.m.