eeg_browse() opens a Shiny app to look through the data. For an
eeg_lst, it shows the signal of the channels and, on request, the
topography of the window, and segments can be selected, for example the
ones to remove. For
an eeg_ica_lst, created with eeg_ica(), it shows
the activations of the components, together with the EOG channels or any
other channels, and the topography of each component, labeled with the
proportion of the variance of the channels it explains (see
eeg_ica_var_tbl()) and its correlation with the EOG channels (see
eeg_ica_cor_tbl()), and components can be selected, for example the ones
to remove.
Usage
eeg_browse(.data, ...)
# S3 method for class 'eeg_lst'
eeg_browse(
.data,
...,
.negative_up = getOption("eeguana.negative_up", FALSE),
.viewer = NULL
)
# S3 method for class 'eeg_ica_lst'
eeg_browse(
.data,
...,
.eog = NULL,
.eog_freq = c(0.1, 30),
.n_components = 16,
.negative_up = getOption("eeguana.negative_up", FALSE),
.viewer = NULL
)Arguments
- .data
An
eeg_lst, or aneeg_ica_lstcreated witheeg_ica().- ...
For an
eeg_lst, the channels shown when the app opens, all of them by default. Not used for aneeg_ica_lst.- .negative_up
Whether negative amplitudes are drawn up, as is usual in some EEG traditions. By default, the value of the option
eeguana.negative_up, which isFALSE, positive up. It can also be changed in the app.- .viewer
Where the app opens, see shiny::viewer. By default, in the browser.
- .eog
Names of the EOG channels, used for the correlations and shown under the components when the app opens. By default, the channels whose names start or end with "eog", ignoring case.
- .eog_freq
Cutoff frequencies in Hz of the filter applied to the EOG channels before correlating them with the components: a band-pass filter from 0.1 to 30 Hz by default. With
NAas the first value, it is only a low-pass filter, and withNAas the second, only a high-pass filter.NULLdoes not filter. It can also be changed in the app.- .n_components
Number of components shown when the app opens, in the order of the variance they explain.
Value
For an eeg_lst, invisibly, the .id of the segments selected. A
message lists them, and shows the calls to eeg_filter() that keep only
them or remove them. When each recording is a single segment, NULL,
invisibly, and no message.
For an eeg_ica_lst, invisibly, a list with one element per recording,
holding the names of the components selected. A message lists them, and
shows the calls to eeg_ica_keep() that keep only them or remove them.
Details
The window can be placed around events and moved from one event to the
next, or it can be moved through the recording. For an eeg_lst, the app
opens moving through the recording, and for an eeg_ica_lst, around events,
when there are any. The window goes on from one segment
into the next, and each segment is shown in its own column. Through the
recording, the start of the window is set relative to the time zero of its
segment, or with a slider over the whole recording, with the segments one
after the other; the window starts where it is asked, and near the end it
stops with the recording. With several
recordings, the buttons "Previous recording" and "Next recording", or the
field, change the recording shown, whose name is in the title. The events are chosen by
their type or their description: the ones that are equal to some values,
start with, end with, or contain some text, or match a regular expression.
For example, the descriptions that start with "peak" are the blinks found
with eeg_artif_peak(). The limits of the window can be set in samples,
milliseconds, or seconds. The left and right arrow keys move to the previous
and next window, and the up and down arrow keys zoom the amplitudes in and
out.
The amplitudes are shown in units of their typical standard deviation: the
median, over stretches of one second of the whole recording, of their
standard deviations, each estimated with the median absolute deviation
(MAD, see stats::mad()), which blinks and other large events barely
change. So a window with a blink and one without are drawn at the same
scale, and slow drifts do not flatten the traces. By default, all the
channels share one scale, and all the components another, so their sizes
can be compared; each trace can also be drawn at its own scale, which makes
the small ones visible. Large amplitudes go on into the rows of the traces
around them, unless the option to cut the traces at the edges of their rows
is on.
The app needs the packages shiny and bslib.
Selecting segments of an eeg_lst
Clicking a segment on the signal selects it, and clicking it again
deselects it; the S key and the "Select segment" button do the same for the segment
of the event, or for the first segment shown. Selected segments are tinted
red. "Done" closes the app and returns the .id of the selected segments;
closing the window returns them as well. What to do with them is up to
eeg_filter(): keeping only them, or removing them. Only segmented data
can be selected: when each recording is a single segment, as before
eeg_segment(), the app is only for browsing, and returns NULL.
The topography is in a sidebar on the right, closed at first, and it is
computed only while it is open. It shows the mean of each channel over the
part of the window in each segment, one head per segment; a box shows one
head for all the segments of the window, their mean, instead. It needs the
positions of the electrodes, see channels_tbl().
Selecting ICA components
Clicking a topography selects the component, and clicking it again
deselects it. "Done" closes the app and returns the selected components;
closing the window returns them as well. What to do with them is up to
eeg_ica_keep(): keeping only them, or removing them.
The EOG channels are filtered before they are correlated with the components, because slow drifts and offsets can hide how closely they follow a component. In the intro vignette, the blink component correlates 0.31 with the unfiltered VEOG and 0.91 with the filtered one. The filter only affects the correlations: the channels are shown unfiltered.
See also
Other ICA functions:
eeg_ica(),
eeg_ica_cor_tbl(),
eeg_ica_keep(),
eeg_ica_show(),
eeg_ica_summary_tbl(),
eeg_ica_var_tbl(),
plot_components()
Other plotting functions:
annotate_electrodes(),
annotate_events(),
annotate_head(),
eeg_downsample(),
ggplot.eeg_lst(),
plot.eeg_lst(),
plot_components(),
plot_in_layout(),
plot_topo(),
theme_eeguana()
Examples
if (interactive()) {
## select the segments to remove
segs <- eeg_segment(data_faces_10_trials, .description %in% c("s70", "s71"),
.lim = c(-.2, .5)
)
to_remove <- eeg_browse(segs)
clean <- eeg_filter(segs, !.id %in% to_remove)
## select the components to remove
ica <- eeg_ica(data_faces_10_trials, -EOGH, -EOGV, -M1, -M2,
.method = fast_ICA
)
to_remove <- eeg_browse(ica)
clean <- eeg_ica_keep(ica, -tidyselect::all_of(to_remove[[1]]))
}
