actChronicFatigue.R

In vincentvanhees/ActChronicFatigue: Process accelerometer data from individuals with chronic fatigue

# Disclaimer: This script is part of the open-source licensed 
# software repository https://github.com/vincentvanhees/ActChronicFatigue
# The developers are not responsible for the use of this software

# Disclaimer: Dit script is onderdeel van de open-source licensed 
# software repository https://github.com/vincentvanhees/ActChronicFatigue
# De ontwikkelaars zijn niet verantwoordelijk voor het gebruik van de software

#===========
#
# Klik op [Source] hier rechtsboven om het programa te starten.
#
rm(list = ls())

gebruik_slaap_dagboek = TRUE
referentiewaarden = c(30,8) # gemiddelde en standaard deviatie
sleeplogidnum = TRUE # TRUE als patient ID een nummer is, FALSE als het ook letters bevat
sep = ";" #","
dec = "," # "."

# hoeveel uren data wil je negeren aan het begin en aan het einde van de meting?
hrs.del.start = 0 # aantal uren te negeren aan het begin
hrs.del.end = 0 # aantal uren te negeren aan het einde


sleepwindowType = "TimeInBed"
# #=========================================
# # Install code if not available:
development.mode = FALSE # laat op FALSE staan, TRUE is alleen voor onderhoud
testbatch = FALSE
if (development.mode == TRUE) {
  roxygen2::roxygenise()
  locationRcode = "D:/Code/ActChronicFatigue/R"
  ffnames = dir(locationRcode) # creating list of filenames of scriptfiles to load
  for (i in 1:length(ffnames)) {
    source(paste(locationRcode, "/", ffnames[i], sep = ""))
  }
} else {
  # cat("\nActChronicFatigue installeren...")
  # remotes::install_github("vincentvanhees/ActChronicFatigue", dependencies = TRUE)
}

if ("GGIR" %in% rownames(installed.packages()) == FALSE) {
  # cat("\nGGIR installeren...")
  # remotes::install_github("wadpac/GGIR", dependencies = TRUE)
}

library(ActChronicFatigue)
# library(GGIR)
library(readxl)
if (development.mode == TRUE) {
  dirR = "D:/Code/GGIR/R"
  ffnames = dir(dirR) # creating list of filenames of scriptfiles to load
  ffnames = ffnames[grep(pattern = "sysdata", x = ffnames, invert = TRUE)]
  for (i in 1:length(ffnames)) {
    source(paste(dirR,"/",ffnames[i], sep = "")) #loading scripts for reading geneactiv data
  }
}

#=========================================
# Obtain data locations
datalocaties = ActChronicFatigue::obtain_folder_paths() # Obtain folder paths from user
replaceslash = function(x) {
  return(gsub(replacement = "/", pattern = "\\\\", x = x))
}
# gt3xdir = replaceslash(datalocaties$gt3xdir)
datadir = replaceslash(datalocaties$datadir)
outputdir = replaceslash(datalocaties$outputdir)
sleeplog = replaceslash(datalocaties$dagboekdir)

cat(paste0(rep('_', options()$width), collapse = ''))
cat("\nBestand locaties:\n")
# cat(paste0("\nLocatie ",length(dir(datalocaties$gt3xdir))," gt3x bestanden = ",datalocaties$gt3xdir))
cat(paste0("\nLocatie ",length(dir(datalocaties$datadir))," gt3x bestanden = ",datalocaties$datadir))
cat(paste0("\nLocatie resultaten =  ",datalocaties$outputdir,"\n"))

chunksize = 0.5
MVPAdefinition = "MVPA_E5S_B10M80._T100_ENMO_0.24hr" #"MVPA_E1M_T100_ENMO_0.24hr"
if (length(dir(datalocaties$datadir)) == 0) { #length(dir(datalocaties$gt3xdir)) == 0 &
  cat("\n")
  stop("\nGeen data gevonden. Controleer data folders.")
}

filenames = dir(datadir, full.names = FALSE)
if (length(filenames) == 0) {
  stop(paste0("\nGeen data bestanden gevonden in ", datadir))
} else {
  cat("\nStart met verwerken van volgende bestand(en):\n")
  print(filenames)
  Q3 = menu(c("Ja", "Nee"), title = paste0("\nDoorgaan?"))
  if (Q3 == 2) {
    stop("\nProgramma gestopt door gebruiker")
  }
}

#=============================================================================
# Start processing of raw accelerometer data with GGIR
cat(paste0(rep('_', options()$width), collapse = ''))
cat("\nStart analyse met GGIR...\n")
ActChronicFatigue::runGGIR(datadir = datadir, outputdir = outputdir, mode = c(1:2),
                           do.report = c(2), overwrite = FALSE,
                           sleepwindowType = "SPT", # to prevent triggering warning
                           visualreport = FALSE, acc.metric = "BFEN", chunksize = chunksize,
                           testbatch = testbatch ,  do.parallel = TRUE, hrs.del.start = hrs.del.start,
                           hrs.del.end = hrs.del.end,
                           sep = sep, dec = dec)
part2resultsfile = paste0(outputdir,"/output_",basename(datadir),"/results/part2_summary.csv")

#=============================================================================
# If sleeplog exists convert sleeplog to expected format
if (gebruik_slaap_dagboek == TRUE) {
  cat("\nConverteer slaapdagboek\n")
  # source("~/projects/ActChronicFatigue/R/convert_sleeplog.R")
  # sleeplogfile = convert_sleeplog(sleeplog, part2resultsfile)
  sleeplogfile = ActChronicFatigue::convert_sleeplog(sleeplog, part2resultsfile, sep = sep)
} else {
  sleeplogfile = c()
}

ActChronicFatigue::runGGIR(datadir = datadir, outputdir = outputdir, mode = c(3:5),
                           do.report = c(4, 5), overwrite = FALSE, do.visual = TRUE,
                           visualreport = FALSE, acc.metric = "BFEN", chunksize = chunksize,
                           loglocation = sleeplogfile, testbatch = testbatch ,
                           do.parallel = TRUE, sleeplogidnum = TRUE, sleepwindowType = sleepwindowType, sep = sep, dec = dec)

#=============================================================================
cat(paste0(rep('_',options()$width), collapse = ''))
cat("\nKlassficeer of de persoon al dan niet laag actief is ...\n")

# Add extra variables, specifically needed for classification
tmp = unlist(strsplit(datadir,"/|\")"))
outputdir2 = outputdir
outputdir2 = paste0(outputdir,"/output_",tmp[length(tmp)])
ActChronicFatigue::addVariables5(outputdir = outputdir2, sep = sep, dec = dec)


# Load the resulting part5_personsummary.csv bestand
part5_summary_file = grep(dir(paste0(outputdir2,"/results"), full.names = TRUE),
                          pattern = "part5_personsummary_WW_", value = T)
part5_summary = read.csv(file = part5_summary_file, sep = sep, dec = dec)

# If all went well this part2_summary object will have a column act90, gradient_mean, towards the end.

# Load the trained model
if (development.mode == TRUE) {
  load("./inst/extdata/final_model_wrist.Rdata")
} else {
  load(system.file("extdata/final_model_wrist.Rdata", package = "ActChronicFatigue"))
}

# Make predictions by applying the model
prop_perv_passive = stats::predict(object = final_model_wrist, newdata = part5_summary, type = "response")
part5_summary = cbind(part5_summary, prop_perv_passive)

CF = coef(final_model_wrist)
model_threshold = -CF[1]/CF[2]

# Note that we assumed that all data in part2_summary comes from accelerometer
# worn on the wrist, because that is what the model was trained for.
# If all went well object prop_perv_passive will have the predictions.

#=============================================================================
# Summarise and show on screen
cat("\n Samenvatting van resultaten:\n")
# source("D:/Code/ActChronicFatigue/R/summarise.R")
ActChronicFatigue::summarise(outputdir2, part5_summary,
                                   model_threshold = model_threshold, referentiewaarden = referentiewaarden,
                                   sleepwindowType = sleepwindowType, MVPAdefinition = MVPAdefinition,
                             threshold_wrist = threshold_wrist, sep = sep, dec = dec)