R/ep.eye_preprocess_gaze.R

In PennStateDEPENdLab/experiment_pipeline: Generates data markdown for NeuroMap Task

#' @title Preprocess gaze data
#' @description Performs AOI tagging (within gaze event fields) and downsampling (within raw gaze data) of gaze data.
#'
#' @param ep.eye Path to a single \code{.edf} file using \code{read_edf()}.
#' @param config config file
#' @param aoi list of AOI options
#' @param task Character value with task name. Contains $indicator, $extract_coords, $extract_labs, $split_coords
#' @param downsample list of downsampling options. Contains $factor, and $method.
#'
#'
#' @return ep.eye an ep.eye object that has gaze data preprocessed.
#'
#' @import dplyr
#' @importFrom data.table data.table
#'
#' @author Nate Hall
#'
#' @export

ep.eye_preprocess_gaze <- function(ep.eye,
                                   config,
                                   aoi,
                                   downsample,
                                   header = NULL){
  ## setup debug
  # ep.eye <- eye_parsed
  # ep.eye <- readRDS("data_proc/eye_preproc/ep.eye_parse_events/595.rds")
  # aoi = config$definitions$eye$gaze_preproc$aoi
  # downsample = config$definitions$eye$gaze_preproc$downsample
  # header = "4. Preprocess gaze data:"

  log_chunk_header(header)

  ### 4.1 Remove impossible values (outside of screen dimensions)
  dt <- "- 4.1 Remove impossible values (outside of screen dimensions):"
  ep.eye <- ep.eye_rm_impossible(ep.eye, dt)

  ### 4.2 Assign AOIs
  if (!is.null(aoi)) {
    cat("- 4.2 Assign AOIs to gaze data\n")
    ep.eye <- ep.eye_add_aois(ep.eye,
                       indicator = aoi$indicator,
                       extract_coords = aoi$extract_coords,
                       extract_labs = aoi$extract_labs,
                       split_coords = aoi$split_coords,
                       tag_raw = aoi$tag_raw)
  } else{
    cat("- 4.2 Assign AOIs to gaze data: SKIP")
  }

  ### 4.3 Collapse timestamps
  dt <- "- 4.3 Collapse timestamps with more than one row:"
  ep.eye <- ep.eye_collapse_time(ep.eye, dt)


  ### N.B. 3/8/21 Interpolating eye data, even in a simplistic manner, is a bad idea given that gaze data is very non-smooth. Better to just rely on the good readings and remove trials with too much data missing, rather than trying to save missing samples.
  ### If the user disagrees the function is still available to use on the preprocessed data, but this should not be included in the standard pipeline.
  # ### 4.4 Interpolate
  # if (!"interpolate" %in% names(c.gaze)) {
  #   cat("- 4.4 Interpolate gaze data: SKIP")
  # } else{
  #   dt <- "- 4.4 Interpolate gaze data:"
  #   tryCatch.ep({
  #     eye <- interpolate_eye(eye, config, signal = "gaze")
  #   },describe_text = dt)
  # }

  ### 4.4 Downsample gaze
  if(!is.null(downsample)){
    tryCatch.ep({
      ep.eye$gaze$downsample <- ep.eye_downsample(ep.eye$raw,
                                                  sample.rate = ep.eye$metadata$sample.rate,
                                                  downsampled_freq = downsample$downsampled_freq,
                                                  analog_channels = c("xp", "yp"),
                                                  method = downsample$method)
    }, describe_text = "- 4.4 Downsample gaze:")
  }

  ### save preproc'ed ep.eye object to correct folder
  if(config$definitions$eye$global$save_steps){
    preproc_dir <- config$definitions$eye$global$preproc_out %>% file.path(., "ep.eye_preproc_gaze")
    if(!dir.exists(preproc_dir)) {dir.create(preproc_dir, recursive = TRUE)}
    subj_path <- file.path(preproc_dir, paste0(config$definitions$eye$global$id, ".rds"))
    tryCatch.ep({
      saveRDS(ep.eye, subj_path)
    },
    describe_text = paste0("- 4.5 Save preprocessed ep.eye [", subj_path,"]:"))
  }


  return(ep.eye)
}