In craddm/eegUtils: Utilities for Electroencephalographic (EEG) Analysis

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  eval = nzchar(Sys.getenv("EVAL_EEG_VIGNETTE"))
)

library(eegUtils)
eeg_example <- import_raw("F:\\Dropbox\\EEGData\\RawEEGData\\BasicCat\\S2B1.bdf",
                          participant_id = "S2")
eeg_example
mod_events <- function(dat) {
  dat <- vapply(seq_along(dat),
                function(i) if (dat[i] == 201) {
                  dat[i] + dat[i - 3]
                  } else {
                    dat[i]}, numeric(1))
  dat
}
eeg_example$events$event_type <- mod_events(eeg_example$events$event_type)
eeg_example <- select_elecs(eeg_example, c("EXG7", "EXG8"), keep = FALSE)
eeg_example <- eeg_reference(eeg_example)
eeg_example <- eeg_filter(eeg_example,
                          low_freq = .1)
eeg_example <- eeg_filter(eeg_example,
                          high_freq = 40)

In this example, we're going to look at how to handle epoching and events. To start, we have continuous data from 70 channels sampled at 512 Hz, average referenced, and bandpass filtered from 0.1 to 40 Hz. There is a complete table of all of the events imported with the data. We can use that table to form epochs.

eeg_example
events(eeg_example)

There are 18 unique trigger codes, which we can check with list_events().

list_events(eeg_example)

In this experiment, six of those triggers correspond to condition codes.

| Trigger | Semantic type | Spatial frequency| |----|--|--| | 207 | Match| High | | 208 | Match| Low| | 213 | Mismatch| High| | 215 | Mismatch| Low| | 219 | Nonsense| High| | 222 | Nonsense| Low|

We can use the epoch_data() function to create epochs around each of these triggers, and label those epochs accordingly. In this example, I'll label each trigger using hierarchically coded event tags, simultaneously giving each condition of semantic type and spatial frequency separated by a "/" symbol.

example_epochs <- 
  epoch_data(eeg_example,
             events = c(207, 
                        208,
                        213,
                        215,
                        219,
                        222),
             epoch_labels = c("Match/HSF",
                              "Match/LSF",
                              "Mismatch/HSF",
                              "Mismatch/LSF",
                              "Nonsense/HSF",
                              "Nonsense/LSF"),
             time_lim = c(-.2, .6))
events(example_epochs)
epochs(example_epochs)

As can be seen, this modifies the epochs structure so that it now labels each epoch accordingly. Since we have hierarchically encoded event tags, an easy way to split that into multiple columns is using the separate() function from the tidyr package.

epochs(example_epochs) <- tidyr::separate(epochs(example_epochs),
                                          col = epoch_labels,
                                          into = c("Semantics", "SF"))
epochs(example_epochs)

Any columns in the epochs structure are included when converting to a data frame. This has a lot of advantages for plotting; many plotting commands now allow you to use facet_wrap() and facet_grid() from ggplot2 to facet on any of the columns from the epoch structure. Note that for performance reasons, facetting must be explicitly allowed using allow_facets = TRUE, since ggplot2's stat_summary() function can be extremely slow.

library(ggplot2)
plot_butterfly(example_epochs,
               legend = FALSE,
               baseline = c(-.1, 0),
               allow_facets = TRUE) +
  facet_wrap(~SF)