Insights from nine nights of self-applied, low-density sleep EEG during sleep restriction therapy: a proof-of-concept evaluation

Code ↔ Paper

7 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 7 matches

[1] § Methods › Outcome Measures › Secondary Outcomes › Power Spectral Analysis ↔ Scripts/RESTORE_Spectral_Analysis_NREM.m, lines 140–187 · score 0.83 · 0.5–32 Hz, FieldTrip, mtmfft, overlap, tapered, beta
[2] § Methods › Outcome Measures › Secondary Outcomes › Power Spectral Analysis ↔ Scripts/RESTORE_Spectral_Analysis_REM.m, lines 140–181 · score 0.75 · 0.5–32 Hz, FieldTrip, mtmfft, overlap, tapered, segments
[3] § Methods › Outcome Measures › Secondary Outcomes › EEG-derived Sleep Architecture ↔ Scripts/06_EEG.Rmd, lines 91–161 · score 0.53 · artefact percentage, blinded, absolute, latency, N1, N2
[4] § Methods › Outcome Measures › Secondary Outcomes › Sleep Diary and Actigraphy-derived Sleep Continuity ↔ Scripts/05_sleep_diary_actigraphy.Rmd, lines 48–152 · score 0.53 · Sleep Diary, survey, Actigraphy, PANAS, waking, SOL
[5] § Results › EEG-Derived Sleep Architecture ↔ Scripts/06_EEG.Rmd, lines 163–245 · score 0.53 · REM latency, sleep stages, N1, N2, N3, SOL
[6] § Methods › Outcome Measures › Statistical Analysis ↔ Scripts/02_follow_up.Rmd, lines 112–156 · score 0.53 · RStudio, change score, Cohen, timepoint, SD, baseline
[7] § Methods › Outcome Measures › Statistical Analysis ↔ Scripts/08_spindles_and_SOs.Rmd, lines 57–101 · score 0.50 · Linear mixed models, LMMs, timepoint, baseline, nights, sleep

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The paper is loaded by your browser, with JavaScript: read it at doi.org or on Europe PMC.

The authors' code

R Markdown · 1,171 lines · 40 KB · CC-BY-4.0 · 2 matches

---
title: "EEG Sleep Analysis"
author: "Emily Stanyer"
date: "`r Sys.Date()`"
output:
html_document:
toc: true
df_print: paged
pdf_document:
toc: true
number_sections: true
params:
# ---- INPUTS ----
profile_dir: "Data/raw/EEG/sleep_profile/" # folder of nightly EEG profile .xlsx files
blinding_file: "Data/raw/EEG/blinding_files/blinding.xlsx" # has ParticipantID, Timepoint, Original_filename, Real_ID (if available)
notes_file: "Data/raw/EEG/scoring_notes/Scoring_Notes_Final_110725.xlsx" # has ParticipantID, artefact_percentage, scoreability, impedances
# ---- OUTPUTS ----
proc_dir: "Data/processed/EEG"
fig_dir: "Figures/EEG"
out_dir: "Output/EEG"
apply_artifact_filter: false # set true to drop nights with artefact % > threshold
artifact_threshold: 10 # percent
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, message = FALSE, warning = FALSE,
fig.width = 10, fig.height = 4.8)
suppressPackageStartupMessages({
library(readr)
library(readxl)
library(dplyr)
library(tidyr)
library(lubridate)
library(stringr)
library(ggplot2)
library(ggprism)
library(writexl)
library(knitr)
library(kableExtra)
library(sessioninfo)
library(lme4)
library(emmeans)
library(purrr)
})
# ---- paths ----
out_dir <- params$out_dir
fig_dir <- params$fig_dir
dir.create(out_dir, recursive = TRUE, showWarnings = FALSE)
dir.create(fig_dir, recursive = TRUE, showWarnings = FALSE)
set.seed(2025)
```
# Read and combine raw EEG profile files
```{r 1-read-eeg}
# Helper: deduplicate readxl-style column suffixes like NAME...1, NAME...2 (keep first)
library(here)
dedup_cols <- function(df){
base <- gsub("\\.{3}\\d+$", "", names(df))
keep <- !duplicated(base)
names(df) <- base
df[, keep, drop = FALSE]
}
# Helper: safer numeric coercion from character
numify <- function(x){ suppressWarnings(as.numeric(x)) }
# 1) List files
# Instead of a hardcoded string, use:
profile_dir <- here::here(params$profile_dir)
stopifnot(dir.exists(profile_dir))
file_list <- list.files(profile_dir, pattern = "\\.xlsx$", full.names = TRUE)
cat("Found", length(file_list), "profile files in", profile_dir, "\n")
if(length(file_list) == 0) stop("No .xlsx files found in ", profile_dir)
# 2) Read each file (skip=1 to match your export), tag ParticipantID from filename (digits before first underscore)
raw_list <- map(file_list, \(pth){
df <- readxl::read_excel(pth, skip = 1, col_types = "text")
df$ParticipantID <- stringr::str_extract(basename(pth), "^\\d+")
dedup_cols(df)
})
raw <- bind_rows(raw_list)
cat("Combined rows:", nrow(raw), " Cols:", ncol(raw), "\n")
```
# Join blinding & artefact/notes; derive day/date/Real_ID
```{r 2-join-blinding-notes}
# --- 1. Define missing helper function ---
load_external <- function(path) {
# Use here::here() to ensure path is absolute relative to project root
abs_path <- here::here(path)
if(!file.exists(abs_path)) stop("External file not found at: ", abs_path)
readxl::read_xlsx(abs_path) %>%
dplyr::mutate(ParticipantID = trimws(as.character(ParticipantID)))
}
# --- 2. Load Blinding and Notes ---
blinding <- load_external(params$blinding_file)
notes <- load_external(params$notes_file)
expected_cols <- c(
"ParticipantID", "Timepoint", "Day",
# Durations
"TIB_DURATION", "SLEEP_DURATION", "WAKE_DURATION", "NONREM_DURATION",
"STAGE1_DURATION", "STAGE2_DURATION", "STAGE3_DURATION", "STAGE4_DURATION",
"REM_DURATION", "WAKE_AFTER_SLEEP_ON", "MOVEMENT_DURATION",
# Percentages
"WAKE_PERCENTAGE", "REM_PERCENTAGE", "NONREM_PERCENTAGE", "SLEEP_EFFICIENCY",
"STAGE1_PERCENTAGE", "STAGE2_PERCENTAGE", "STAGE3_PERCENTAGE", "DEEPSLEEP_PERCENTAGE",
# Latencies
"SLEEPLATENCY", "REM_LATENCY", "SLEEPLATENCY_DEEPSLEEP", "SLEEPLATENCY_LPS",
# Other Metrics
"FIRST_REM", "STAGE_CHANGE_NUMBER", "WAKE_NUMBER", "ARTEFACT_PERCENTAGE",
"REM_DENSITY", "SLEEP_PROFILE_MANUALLY_SCORED"
)
# --- 3. Join and Derive ---
merged <- raw %>%
dplyr::mutate(ParticipantID = trimws(as.character(ParticipantID))) %>%
# Ensure expected_cols contains all the software headers we discussed!
dplyr::select(dplyr::any_of(c(expected_cols, "ParticipantID", "Original_filename", "Timepoint"))) %>%
dplyr::left_join(blinding, by = "ParticipantID", suffix = c("", ".bl")) %>%
dplyr::left_join(notes %>% dplyr::select(ParticipantID, dplyr::any_of(c("artefact_percentage", "scoreability"))),
by = "ParticipantID") %>%
dplyr::mutate(
# 1. Handle filename and ID derivation
Original_filename = dplyr::coalesce(
!!!rlang::syms(intersect(names(.), c("Original_filename", "Original_filename.x", "Original_filename.bl"))),
ParticipantID
),
Real_ID = dplyr::coalesce(
!!!rlang::syms(intersect(names(.), c("Real_ID", "Real_ID.bl", "Real_ID.y"))),
stringr::str_extract(Original_filename, "(?<=REST_)\\d{3}"),
ParticipantID
),
# 2. Extract Date
date_str = stringr::str_extract(Original_filename, "\\d{8}"),
date = suppressWarnings(lubridate::dmy(date_str)),
# 3. IMPROVED Day Extraction (Fixes the NA issue)
# This looks for 'N' or 'NIGHT' followed by the number
day_num = stringr::str_extract(Original_filename, "(?i)(?<=NIGHT|N|N_)\\d+"),
day = dplyr::case_when(
stringr::str_detect(Original_filename, "(?i)BSL|Baseline") ~ "BSL",
stringr::str_detect(Original_filename, "(?i)FU|Follow") ~ "FU",
!is.na(day_num) ~ as.character(day_num), # Maps 'N1' to '1', 'N2' to '2', etc.
TRUE ~ as.character(Timepoint) # Fallback to Timepoint column
)
) %>%
dplyr::select(-dplyr::any_of(c("Original_filename.x","Original_filename.y","Original_filename.bl",
"Real_ID.x","Real_ID.y","Real_ID.bl", "date_str", "day_num")))
```
# Clean the data
```{r 3-clean-transform, message=FALSE, warning=FALSE}
# 1. Settings
day_levels <- c("BSL","1","2","3","4","5","6","7","FU")
# Heuristic unit converter: day-fraction -> minutes; hours -> minutes; else assume minutes
convert_to_minutes <- function(x){
x <- numify(x)
if (all(is.na(x))) return(x)
xp <- x[!is.na(x)]
if (length(xp) > 0 && max(xp, na.rm = TRUE) <= 1.1) {
x * 24 * 60
} else if (length(xp) > 0 && stats::median(xp, na.rm = TRUE) <= 20) {
x * 60
} else x
}
# Columns that represent durations (will be converted to minutes)
duration_vars <- c(
"TIB_DURATION","SLEEP_DURATION","WAKE_DURATION","REM_DURATION","NONREM_DURATION",
"REM_LATENCY","SLEEPLATENCY","SLEEPLATENCY_NONREM","WAKE_AFTER_SLEEP_ON",
"STAGE1_DURATION","STAGE2_DURATION","STAGE3_DURATION","STAGE4_DURATION",
"SLEEPLATENCY_DEEPSLEEP","SLEEPLATENCY_LPS","MOVEMENT_DURATION",
"MOVEMENT_DURATION_SPT","WAKE_DURATION_SPT","REM_DURATION_SPT",
"STAGE1_DURATION_SPT","STAGE2_DURATION_SPT","STAGE3_DURATION_SPT","STAGE4_DURATION_SPT",
"ARTEFACT_DURATION_SPT","FIRST_REM","REM_LATENCY_MSLT","REM_LATENCY_MWT"
)
# 2. Clean the raw column names
# Removes Excel suffixes like "...15" and trims hidden spaces
names(merged) <- trimws(gsub("\\.{3}\\d+$", "", names(merged)))
# 3. Define the Rename Map (NEW NAME = OLD SOFTWARE NAME)
rename_map <- c(
"Wake (mins)" = "WAKE_DURATION",
"REM (mins)" = "REM_DURATION",
"N1 (mins)" = "STAGE1_DURATION",
"N2 (mins)" = "STAGE2_DURATION",
"N3 (mins)" = "STAGE3_DURATION",
"Wake (%)" = "WAKE_PERCENTAGE",
"REM (%)" = "REM_PERCENTAGE",
"N1 (%)" = "STAGE1_PERCENTAGE",
"N2 (%)" = "STAGE2_PERCENTAGE",
"N3 (%)" = "STAGE3_PERCENTAGE",
"WASO (mins)" = "WAKE_AFTER_SLEEP_ON",
"REM Latency (mins)" = "REM_LATENCY",
"SOL (mins)" = "SLEEPLATENCY",
"TIB (mins)" = "TIB_DURATION",
"TST (mins)" = "SLEEP_DURATION",
"Sleep Efficiency (%)" = "SLEEP_EFFICIENCY",
"N3 Latency (mins)" = "SLEEPLATENCY_DEEPSLEEP",
"Sleep Stage Transitions" = "STAGE_CHANGE_NUMBER",
"Transitions to Wake" = "WAKE_NUMBER")
# 4. Build the 'clean' dataframe
clean <- merged %>%
# Convert durations (using the helper function defined earlier)
dplyr::mutate(dplyr::across(dplyr::any_of(duration_vars), convert_to_minutes)) %>%
# Apply the renaming map
dplyr::rename(dplyr::any_of(rename_map)) %>%
# Ensure correct data types and factors
dplyr::mutate(
Real_ID = as.character(Real_ID),
day = factor(day, levels = day_levels),
# Ensure these are numeric even if Excel had them as text
artefact_percentage = suppressWarnings(as.numeric(as.character(artefact_percentage))),
scoreability = suppressWarnings(as.numeric(as.character(scoreability)))
)
# --- NOTE: Participant and Artefact filtering removed per request ---
final_output_path <- here::here("Data", "processed", "EEG")
# 5. Final Export of the Wide Table
writexl::write_xlsx(clean, file.path(final_output_path, "clean_EEG_data.xlsx"))
# --- DIAGNOSTICS ---
cat("--- Data Clean Check ---\n")
cat("Total rows kept:", nrow(clean), "\n")
cat("Total participants:", dplyr::n_distinct(clean$Real_ID), "\n")
cat("Columns renamed successfully:", sum(names(rename_map) %in% names(clean)), "out of", length(rename_map), "\n")
if(any(!names(rename_map) %in% names(clean))) {
cat("Missing from raw data:", paste(names(rename_map)[!names(rename_map) %in% names(clean)], collapse=", "), "\n")
}
```
# Long data + exports
```{r 4-long-data}
# 1. Define the variables we want to keep for analysis/plotting
# (Must match the NEW names from Chunk 3 exactly)
needed_vars <- c(
"TIB (mins)", "TST (mins)", "Wake (mins)", "Wake (%)",
"REM (mins)", "REM (%)", "N1 (mins)", "N1 (%)",
"N2 (mins)", "N2 (%)", "N3 (mins)", "N3 (%)",
"WASO (mins)", "REM Latency (mins)", "SOL (mins)",
"Sleep Efficiency (%)", "N3 Latency (mins)",
"Sleep Stage Transitions", "Transitions to Wake", # Added these
"artefact_percentage"
)
# 2. Pivot to Long Format
EEG_long <- clean %>%
# Select ID, Day, and only the columns in our 'needed_vars' list
dplyr::select(Real_ID, day, dplyr::any_of(needed_vars)) %>%
# Ensure all value columns are numeric before pivoting
# (Prevents "Can't combine double and character" errors)
dplyr::mutate(dplyr::across(-c(Real_ID, day), ~suppressWarnings(as.numeric(as.character(.x))))) %>%
# Transform from Wide to Long
tidyr::pivot_longer(
cols = -c(Real_ID, day),
names_to = "variable",
values_to = "value"
) %>%
# Remove any rows where the value is NA (optional, helps keep the file small)
dplyr::filter(!is.na(value))
# 3. Export to Excel
# --- NOTE: Participant and Artefact filtering removed per request ---
final_output_path <- here::here("Data", "processed", "EEG")
# 5. Final Export of the Wide Table
writexl::write_xlsx(clean, file.path(final_output_path, "EEG_data_long_clean.xlsx"))
# --- DIAGNOSTICS ---
cat("--- Long Data Check ---\n")
cat("Total rows in long format:", nrow(EEG_long), "\n")
cat("Variables found and pivoted:\n")
print(unique(EEG_long$variable))
# Quick Peek at the first few rows
head(EEG_long)
```
# Summary statistics (Mean, SD, CI) and wide tables
```{r summaries}
# SAFETY CHECK: If EEG_long doesn't exist or is empty, stop and warn
if (!exists("EEG_long") || nrow(EEG_long) == 0) {
stop("EEG_long is missing or empty. Please run the 'long-data' chunk first.")
}
# 1) Calculate Statistics
sum_long <- EEG_long %>%
# Filter out rows where value is NA to avoid math errors
filter(!is.na(value)) %>%
group_by(variable, day) %>%
summarise(
n = n(),
mean = mean(value, na.rm = TRUE),
sd = sd(value, na.rm = TRUE),
se = sd / sqrt(pmax(n, 1)),
tcrit = qt(0.975, df = pmax(n-1, 1)),
ci_low = mean - tcrit * se,
ci_high = mean + tcrit * se,
.groups = "drop"
) %>%
# Format into strings for the table
mutate(
`Mean (SD)` = sprintf("%.2f (%.2f)", mean, sd),
CI = paste0(sprintf("%.2f", ci_low), "–", sprintf("%.2f", ci_high))
)
# 2) Create Wide Tables
# Table 1: Mean (SD)
mean_sd_wide <- sum_long %>%
select(variable, day, `Mean (SD)`) %>%
pivot_wider(names_from = day, values_from = `Mean (SD)`) %>%
arrange(variable)
# Table 2: N (Sample Size)
n_wide <- sum_long %>%
select(variable, day, n) %>%
pivot_wider(names_from = day, values_from = n) %>%
arrange(variable)
output_path <- here::here("Output", "EEG")
# 4) Display in the Report
cap <- "EEG Mean (SD) by Day (Rows = Variables; Columns = BSL, 1–7, FU)"
knitr::kable(mean_sd_wide, caption = cap) %>%
kableExtra::kable_styling(full_width = FALSE, bootstrap_options = c("striped", "hover", "condensed"))
```
# Plots (overall and minutes-only)
```{r plots, fig.width=11, fig.height=9}
library(ggprism)
# 1. Prepare data for plotting
plot_df <- EEG_long %>%
group_by(day, variable) %>%
summarise(
mean = mean(value, na.rm = TRUE),
sd = sd(value, na.rm = TRUE),
n = sum(!is.na(value)),
se = sd / sqrt(pmax(n, 1)),
.groups = "drop"
)
# 2. Define the selection (Minutes + Sleep Efficiency)
# Adding "Sleep Efficiency (%)" to your list
mins_plus_se <- c(
"TIB (mins)", "TST (mins)", "Sleep Efficiency (%)",
"WASO (mins)", "SOL (mins)", "REM Latency (mins)",
"N3 Latency (mins)", "N1 (mins)", "N2 (mins)",
"N3 (mins)", "REM (mins)"
)
# 3. Filter and set factor levels for stable plotting order
plot_filtered <- plot_df %>%
dplyr::filter(variable %in% mins_plus_se) %>%
dplyr::mutate(variable = factor(variable, levels = mins_plus_se))
# 4. Generate the Plot (Fixed Theme Order)
p_eeg <- ggplot(plot_filtered, aes(x = day, y = mean, group = 1)) +
geom_point(size = 3, shape = 17, color = "#0091d4") +
geom_line(linewidth = 1, color = "#0091d4") +
geom_errorbar(aes(ymin = mean - se, ymax = mean + se), width = 0.2, color = "#0091d4") +
# 1. Faceting
facet_wrap(~ variable, scales = "free_y", ncol = 3) +
# 3. Custom Overrides (Must come AFTER theme_prism)
theme(
legend.position = "none",
strip.text = element_text(size = 9),
axis.title = element_text(face = "bold"),
plot.title = element_text(hjust = 0.5)
) +
# 4. Labels
labs(
x = "Study Night",
y = "Mean ± SE",
)
# 4. Generate the Plot (Clean Publication Style)
p_eeg <- ggplot(plot_filtered, aes(x = day, y = mean, group = 1)) +
# Geoms
geom_line(linewidth = 1, color = "#0091d4") +
geom_errorbar(aes(ymin = mean - se, ymax = mean + se), width = 0.2, color = "#0091d4") +
geom_point(size = 3, shape = 17, color = "#0091d4") +
# Faceting - "free_y" allows each metric to have its own scale
facet_wrap(~ variable, scales = "free_y", ncol = 3, axes = "all") +
# --- MANUAL CLEAN THEME ---
theme_classic() + # Removes grid and gray background automatically
theme(
# 1. Force Black Axis Lines (Prism Style)
axis.line = element_line(colour = "black", linewidth = 0.8),
axis.ticks = element_line(colour = "black", linewidth = 0.8),
# 2. Format Text
axis.text = element_text(color = "black", size = 10),
axis.title = element_text(face = "bold", size = 11),
# 3. Clean Facet Labels (No gray boxes)
strip.background = element_blank(),
strip.text = element_text(size = 10, face = "bold"),
# 4. Final Polish
legend.position = "none",
panel.spacing = unit(1, "lines") # Adds a bit of air between plots
) +
# --- END THEME ---
labs(
x = "Study Night",
y = "Mean ± SE"
)
# Display
print(p_eeg)
fig_dir <- here::here("Figures", "EEG")
# Create the folder if it doesn't exist
if (!dir.exists(fig_dir)) {
dir.create(fig_dir, recursive = TRUE)
}
# Save the high-res TIFF
ggsave(file.path(fig_dir, "Figure 4.tiff"), p_eeg, width = 10, height = 7, dpi = 300, compression = "lzw")
cat("Plot successfully saved to:", file.path(fig_dir, "Figure_4.tiff"))
# 5. Save Exports
ggsave(file.path(fig_dir, "Figure 4.jpg"), p_eeg, width = 10, height = 7, dpi = 300)
```
# LMM function
```{r}
run_eeg_lmm <- function(df_long, varname) {
# 1. Prepare subset and ensure Value is numeric
df_sub <- df_long %>%
filter(variable == varname) %>%
mutate(
day = factor(day),
value = as.numeric(as.character(value))
) %>%
filter(!is.na(value))
# 2. Check levels
levels_present <- unique(as.character(df_sub$day))
if (!("BSL" %in% levels_present) | length(levels_present) < 2) return(NULL)
# 3. Fit Model
fit <- tryCatch({
lmer(value ~ day + (1 | Real_ID), data = df_sub, REML = TRUE)
}, error = function(e) return(NULL))
if (is.null(fit)) return(NULL)
# 4. emmeans
em <- emmeans(fit, "day")
ct <- contrast(em, method = "trt.vs.ctrl", ref = "BSL") %>%
summary(infer = TRUE) %>%
as.data.frame()
# 5. Calculate Cohen's d (Safety update here)
# Pivot and ensure columns are numeric for the subtraction
wide_data <- df_sub %>%
select(Real_ID, day, value) %>%
pivot_wider(names_from = day, values_from = value) %>%
mutate(across(-Real_ID, ~as.numeric(as.character(.x)))) # FORCE NUMERIC
days_to_compare <- setdiff(levels_present, "BSL")
d_results <- map_dfr(days_to_compare, function(d_val) {
# Check if both columns exist and are numeric
if ("BSL" %in% names(wide_data) && d_val %in% names(wide_data)) {
# Use na.rm = TRUE to handle missing values within the pair
diffs <- wide_data[[d_val]] - wide_data$BSL
mean_diff <- mean(diffs, na.rm = TRUE)
sd_diff <- sd(diffs, na.rm = TRUE)
cohens_d <- if(!is.na(sd_diff) && sd_diff != 0) mean_diff / sd_diff else NA
return(tibble(contrast = paste0(d_val, " - BSL"), Cohens_d = cohens_d))
}
return(NULL)
})
# 6. Combine
res <- ct %>%
left_join(d_results, by = "contrast") %>%
mutate(variable = varname)
return(res)
}
```
# Linear mixed models (contrasts vs BSL) with within-person d
```{r}
suppressPackageStartupMessages({
library(flextable)
library(officer)
library(dplyr)
library(tidyr)
})
# 1. Prepare Data
word_table_df <- sum_long %>%
filter(variable %in% target_vars) %>%
mutate(contrast = paste0(day, " - BSL")) %>%
left_join(EEG_effects, by = c("variable", "contrast")) %>%
mutate(
LCL = dplyr::coalesce(!!!rlang::syms(intersect(names(.), c("lower.CL", "asymp.LCL")))),
UCL = dplyr::coalesce(!!!rlang::syms(intersect(names(.), c("upper.CL", "asymp.UCL"))))
) %>%
mutate(variable = factor(variable, levels = target_vars)) %>%
arrange(variable, day) %>%
mutate(
`Mean (SD)` = sprintf("%.2f (%.2f)", mean, sd),
`Diffadj` = ifelse(day == "BSL", "-", sprintf("%.2f", estimate)),
`95% CI` = ifelse(day == "BSL" | is.na(LCL), "-",
paste0(sprintf("%.2f", LCL), " to ", sprintf("%.2f", UCL))),
`d` = ifelse(day == "BSL" | is.na(Cohens_d), "-", sprintf("%.2f", Cohens_d)),
Day_Label = dplyr::case_when(
day == "BSL" ~ "Baseline",
day == "FU" ~ "Follow up",
TRUE ~ paste("Night", day)
)
) %>%
select(variable, Day_Label, `Mean (SD)`, `Diffadj`, `95% CI`, `d`)
# 2. Convert to Grouped Data
grouped_data <- as_grouped_data(x = word_table_df, groups = c("variable"))
# Calculate group IDs for shading (ignoring the header rows)
grouped_data <- grouped_data %>%
mutate(
temp_var = if_else(!is.na(variable), as.character(variable), NA_character_)
) %>%
fill(temp_var, .direction = "down") %>%
mutate(group_num = as.numeric(factor(temp_var, levels = target_vars)))
# 3. Create Flextable
ft <- flextable(grouped_data, col_keys = c("variable", "Day_Label", "Mean (SD)", "Diffadj", "95% CI", "d")) %>%
# Header labels
set_header_labels(
variable = "EEG (n = 17)",
Day_Label = "",
`Mean (SD)` = "Mean (SD)",
Diffadj = "Diffadj",
`95% CI` = "95% CI",
d = "d"
) %>%
# Remove Italics
italic(italic = FALSE, part = "all") %>%
# Bold variable headers
bold(i = ~ !is.na(variable), j = 1, bold = TRUE) %>%
# Apply Shading (even groups get grey)
bg(i = ~ group_num %% 2 == 0, bg = "#F2F2F2", part = "body") %>%
# FIX: Hide the variable names in data rows using compose
compose(i = ~ is.na(variable), j = "variable", value = as_paragraph("")) %>%
# Styling
fontsize(size = 10, part = "all") %>%
align(align = "left", part = "all") %>%
autofit() %>%
# Borders
border_remove() %>%
hline_top(border = fp_border(width = 1.5), part = "header") %>%
hline(i = 1, border = fp_border(width = 1.5), part = "header") %>%
# Footer
add_footer_lines("95% CI, 95% confidence interval of the mean difference; Diffadj, mean difference adjusted for missing values derived from linear mixed model; d, effect size (Cohen’s d); SD, standard deviation.") %>%
hline_bottom(border = fp_border(width = 1.5), part = "footer")
# 4. Export
doc <- read_docx() %>%
body_add_flextable(ft)
word_out <- here::here("Output", "EEG", "EEG_LMM_results.docx")
print(doc, target = word_out)
cat("Success! Table saved to:", word_out)
```
# LMM for relative sleep stages for supplementary
```{r}
# --- 1. Define Percentage Variables ---
target_perc_vars <- c("N1 (%)", "N2 (%)", "N3 (%)", "REM (%)")
# --- 2. Run the LMM Logic for Percentages ---
res_list_perc <- lapply(target_perc_vars, \(v) if(v %in% unique(EEG_long$variable)) run_eeg_lmm(EEG_long, v) else NULL)
EEG_effects_perc <- dplyr::bind_rows(purrr::compact(res_list_perc))
# --- 3. Join and Format Data ---
final_perc_table_df <- sum_long %>%
filter(variable %in% target_perc_vars) %>%
mutate(contrast = paste0(day, " - BSL")) %>%
left_join(EEG_effects_perc, by = c("variable", "contrast")) %>%
# Fix for emmeans column naming
mutate(
LCL = dplyr::coalesce(!!!rlang::syms(intersect(names(.), c("lower.CL", "asymp.LCL")))),
UCL = dplyr::coalesce(!!!rlang::syms(intersect(names(.), c("upper.CL", "asymp.UCL"))))
) %>%
mutate(variable = factor(variable, levels = target_perc_vars)) %>%
arrange(variable, day) %>%
mutate(
`Mean (SD)` = sprintf("%.2f (%.2f)", mean, sd),
`Diffadj` = ifelse(day == "BSL", "-", sprintf("%.2f", estimate)),
`95% CI` = ifelse(day == "BSL" | is.na(LCL), "-",
paste0(sprintf("%.2f", LCL), " to ", sprintf("%.2f", UCL))),
`d` = ifelse(day == "BSL" | is.na(Cohens_d), "-", sprintf("%.2f", Cohens_d)),
Day_Label = dplyr::case_when(
day == "BSL" ~ "Baseline",
day == "FU" ~ "Follow up",
TRUE ~ paste("Night", day)
)
) %>%
select(variable, Day_Label, `Mean (SD)`, `Diffadj`, `95% CI`, `d`)
# --- 4. Create the Formatted Word Table ---
grouped_perc_data <- as_grouped_data(x = final_perc_table_df, groups = c("variable"))
# Shading logic
grouped_perc_data <- grouped_perc_data %>%
mutate(temp_var = if_else(!is.na(variable), as.character(variable), NA_character_)) %>%
fill(temp_var, .direction = "down") %>%
mutate(group_num = as.numeric(factor(temp_var, levels = target_perc_vars)))
ft_perc <- flextable(grouped_perc_data, col_keys = c("variable", "Day_Label", "Mean (SD)", "Diffadj", "95% CI", "d")) %>%
set_header_labels(variable = "EEG % (n = 17)", Day_Label = "", `Mean (SD)` = "Mean (SD)",
Diffadj = "Diffadj", `95% CI` = "95% CI", d = "d") %>%
italic(italic = FALSE, part = "all") %>%
bold(i = ~ !is.na(variable), j = 1, bold = TRUE) %>%
bg(i = ~ group_num %% 2 == 0, bg = "#F2F2F2", part = "body") %>%
compose(i = ~ is.na(variable), j = "variable", value = as_paragraph("")) %>%
fontsize(size = 10, part = "all") %>%
align(align = "left", part = "all") %>%
autofit() %>%
border_remove() %>%
hline_top(border = fp_border(width = 1.5), part = "header") %>%
hline(i = 1, border = fp_border(width = 1.5), part = "header") %>%
add_footer_lines("95% CI, 95% confidence interval; Diffadj, mean difference adjusted via LMM; d, Cohen’s d; SD, standard deviation.") %>%
hline_bottom(border = fp_border(width = 1.5), part = "footer")
# --- 5. Export ---
doc_perc <- read_docx() %>% body_add_flextable(ft_perc)
word_perc_out <- here::here("Output", "EEG", "EEG_LMM_Results_Proportions_supplementary.docx")
print(doc_perc, target = word_perc_out)
cat("Percentage results saved to Word:", word_perc_out)
```
# Create forest plots to show effect sizes
```{r, fig.width=6, fig.height=9}
library(dplyr)
library(stringr)
library(ggplot2)
# 1. Define the chronological order for the Y-axis (Top to Bottom)
night_levels <- c("1", "2", "3", "4", "5", "6", "7", "FU")
night_labels <- c("Night 1", "Night 2", "Night 3", "Night 4", "Night 5", "Night 6", "Night 7", "Follow up")
# 2. Prepare the data
plot_data_final <- EEG_effects %>%
# Filter out any rows where LMM didn't run
filter(!is.na(estimate)) %>%
mutate(
# FIX: Create the 'day' column by extracting word/number before the dash
# This turns "1 - BSL" into "1" and "FU - BSL" into "FU"
extracted_day = str_extract(contrast, "^[^ ]+"),
# Now create the Day_Label factor for the Y-axis
Day_Label = factor(extracted_day,
levels = rev(night_levels),
labels = rev(night_labels)),
# Ensure math columns are numeric
estimate = as.numeric(as.character(estimate)),
low = as.numeric(lower.CL),
high = as.numeric(upper.CL),
abs_d = abs(as.numeric(as.character(Cohens_d))),
# Cohen's d sizing bins
d_size = case_when(
abs_d < 0.2 ~ "Negligible (< 0.2)",
abs_d >= 0.2 & abs_d < 0.5 ~ "Small (0.2)",
abs_d >= 0.5 & abs_d < 0.8 ~ "Medium (0.5)",
abs_d >= 0.8 ~ "Large (0.8)",
TRUE ~ "Small (0.2)"
),
d_size = factor(d_size, levels = c("Negligible (< 0.2)", "Small (0.2)", "Medium (0.5)", "Large (0.8)"))
) %>%
filter(!is.na(Day_Label)) # Drops anything that didn't match our 1-7/FU list
# 3. Split into your two requested groups
df_summary_plot <- plot_data_final %>%
filter(variable %in% c("TST (mins)", "TIB (mins)", "WASO (mins)", "SOL (mins)", "Sleep Efficiency (%)"))
df_stages_plot <- plot_data_final %>%
filter(variable %in% c("N1 (mins)", "N2 (mins)", "N3 (mins)", "REM (mins)"))
library(ggplot2)
make_final_eeg_plot <- function(data_subset, plot_title, theme_color) {
# 1. Symmetric limits for the 0-anchor
limit_val <- max(abs(c(data_subset$low, data_subset$high)), na.rm = TRUE) * 1.1
if(is.infinite(limit_val) || is.na(limit_val)) limit_val <- 10
ggplot(data_subset, aes(x = estimate, y = Day_Label)) +
# Reference line at zero
geom_vline(xintercept = 0, linetype = "dashed", color = "gray70") +
# Error bars matching theme_color
geom_errorbarh(aes(xmin = low, xmax = high),
height = 0.3, linewidth = 0.8, color = theme_color, alpha = 0.5) +
# Blobs (Points) sized by Cohen's d
geom_point(aes(size = d_size), color = theme_color) +
# Separate plots for each metric
facet_wrap(~variable, scales = "free_x", ncol = 2) +
expand_limits(x = c(-limit_val, limit_val)) +
# Define the "Blob" sizes based on Cohen's conventions
scale_size_manual(values = c(
"Negligible (< 0.2)" = 1.5,
"Small (0.2)" = 3,
"Medium (0.5)" = 5,
"Large (0.8)" = 8
)) +
labs(
title = plot_title,
subtitle = "(n = 17)",
x = "Adjusted Mean Difference (95% CI)",
y = NULL,
size = "Effect Size (d)"
) +
theme_minimal() +
theme(
strip.text = element_text(face = "bold", size = 9),
panel.grid.minor = element_blank(),
panel.spacing = unit(1.2, "lines"),
legend.position = "bottom"
)
}
# 4. Generate the Plots
p_summary <- make_final_eeg_plot(df_summary_plot, "EEG Sleep Continuity", "#1B4F72")
p_stages <- make_final_eeg_plot(df_stages_plot, "EEG Sleep Architecture", "#0E6251")
# 5. Display
print(p_summary)
# 6. Save the Plots
ggsave(file.path(fig_dir, "Sleep_continuity.jpg"), p_summary, width = 6, height = 9, dpi = 600)
print(p_stages)
ggsave(file.path(fig_dir, "Sleep_stages.jpg"), p_stages, width = 6, height = 7, dpi = 600)
```
# Create heat map of sleep stage changes
```{r}
library(tidyverse)
library(lubridate)
# 1. PATH SETUP (Now platform-independent)
input_path <- here::here("Data", "raw", "EEG", "hypnograms_txt")
# 2. Verify and Load
if (!dir.exists(input_path)) {
stop("Path not found! Check if 'Data/raw/EEG/hypnograms_txt' exists in your Project folder.")
}
all_files <- list.files(input_path, pattern = "\\.txt$", full.names = TRUE, ignore.case = TRUE)
cat("Success! Found", length(all_files), "files in:", input_path)
all_files <- list.files(input_path, pattern = "\\.txt$", full.names = TRUE)
# --- 1. UPDATED LOADER (Ensures Real_ID exists) ---
read_hypno_by_date <- function(f) {
fname <- basename(f)
# Extract ID (First 3 digits) and Date (YYYY_MM_DD)
subj_id <- stringr::str_extract(fname, "^\\d{3}")
date_str <- stringr::str_extract(fname, "\\d{4}_\\d{2}_\\d{2}")
# Skip file if naming doesn't match expected pattern
if(is.na(subj_id) | is.na(date_str)) return(NULL)
# Read the numeric data
df <- tryCatch({
read.table(f, header = FALSE, sep = "")
}, error = function(e) return(NULL))
if(is.null(df) || ncol(df) < 1) return(NULL)
# Standardize and add metadata
df %>%
rename(stage = V1, exclude = V2) %>%
filter(stage != 6, exclude == 0) %>%
mutate(
stage_name = case_when(
stage == 0 ~ "Wake", stage == 1 ~ "N1", stage == 2 ~ "N2",
stage == 3 ~ "N3", stage %in% c(4, 5) ~ "REM", TRUE ~ NA_character_
),
# ENSURE THESE COLUMNS ARE CREATED FOR EVERY ROW
Real_ID = paste0("REST_", subj_id),
file_date = lubridate::ymd(date_str),
is_bsl = str_detect(fname, "(?i)Bsl|Baseline")
)
}
# --- 2. RUN THE LOADING ---
raw_master <- purrr::map_df(all_files, read_hypno_by_date)
# Check if Real_ID actually exists now
if(!"Real_ID" %in% names(raw_master)) stop("Real_ID column missing after load!")
# --- 3. UNBLINDING AND LABELING ---
unblinded_master <- raw_master %>%
group_by(Real_ID) %>%
# Ensure chronological order per participant
arrange(file_date, .by_group = TRUE) %>%
mutate(
day_type = if_else(is_bsl, "BSL", "Intervention")
) %>%
# Now, within each participant, rank the 'Intervention' nights
group_by(Real_ID, day_type) %>%
mutate(
night_num = if_else(day_type == "Intervention", dense_rank(file_date), NA_integer_)
) %>%
ungroup() %>%
mutate(
# Final Labeling
day_label = case_when(
day_type == "BSL" ~ "BSL",
night_num == 8 ~ "FU", # Logic: If 8th night after BSL, it's Follow Up
!is.na(night_num) ~ paste0("Night ", night_num),
TRUE ~ "Other"
),
day_label = factor(day_label, levels = c("BSL", paste0("Night ", 1:7), "FU"))
)
# --- 4. CALCULATE TRANSITIONS ---
stage_order <- c("Wake", "N1", "N2", "N3", "REM")
final_probs <- unblinded_master %>%
group_by(Real_ID, day_label) %>% # Calculate transitions WITHIN each night
mutate(
current = factor(stage_name, levels = stage_order),
next_s = factor(lead(stage_name), levels = stage_order)
) %>%
filter(!is.na(current), !is.na(next_s)) %>%
group_by(day_label, current, next_s) %>%
summarise(count = n(), .groups = 'drop_last') %>%
mutate(prob = (count / sum(count)) * 100) %>%
ungroup()
bsl_vals <- final_probs %>%
filter(day_label == "BSL") %>%
select(current, next_s, bsl_prob = prob)
diff_matrix <- final_probs %>%
filter(day_label != "BSL") %>%
left_join(bsl_vals, by = c("current", "next_s")) %>%
mutate(
bsl_prob = replace_na(bsl_prob, 0),
diff = prob - bsl_prob
)
# --- 3. GENERATE THE ORDERED HEATMAP ---
ggplot(diff_matrix, aes(x = current, y = next_s, fill = diff)) +
geom_tile(color = "white", linewidth = 0.3) +
geom_text(aes(label = sprintf("%.1f%%", diff)),
size = 2.5, fontface = "bold") +
scale_fill_gradient2(
low = "#4a1486", mid = "white", high = "#d95f0e",
midpoint = 0, name = "Δ from BSL %"
) +
facet_wrap(~day_label, ncol = 4) +
theme_minimal() +
labs(
x = "From Stage (Epoch n)",
y = "To Stage (Epoch n+1)"
) +
theme(
axis.text.x = element_text(angle = 45, hjust = 1),
strip.text = element_text(face = "bold")
)
library(ggplot2)
# 1. DEFINE FILENAME AND PATH
# Use here() to point specifically to your Figures/EEG folder
output_path <- here::here("Figures", "EEG", "Figure S3.jpg")
# 2. SAVE THE PLOT
ggsave(
filename = output_path,
plot = last_plot(),
device = "jpeg", # Changed to 'jpeg' for consistency with .jpg extension
dpi = 300,
width = 12,
height = 6,
units = "in"
)
message("Success! Plot saved to: ", output_path)
```
# Adaptation nights
```{r, fig.width=12, fig.height=20, warning=FALSE, message=FALSE}}
# --- 1. SETUP HELPERS & MAPPINGS ---
# Define the bridge between Adaptation (CAPS) and Clean (Pretty) names
rename_map <- c(
"TIB (mins)" = "TIB_DURATION",
"TST (mins)" = "SLEEP_DURATION",
"Wake (mins)" = "WAKE_DURATION",
"N1 (mins)" = "STAGE1_DURATION",
"N2 (mins)" = "STAGE2_DURATION",
"N3 (mins)" = "STAGE3_DURATION",
"REM (mins)" = "REM_DURATION",
"WASO (mins)" = "WAKE_AFTER_SLEEP_ON",
"Sleep Efficiency (%)" = "SLEEP_EFFICIENCY",
"SOL (mins)" = "SLEEPLATENCY",
"REM Latency (mins)" = "REM_LATENCY",
"N3 Latency (mins)" = "SLEEPLATENCY_DEEPSLEEP"
)
# List of duration variables to be converted to minutes
duration_vars <- c("TIB_DURATION", "SLEEP_DURATION", "WAKE_DURATION",
"STAGE1_DURATION", "STAGE2_DURATION", "STAGE3_DURATION",
"REM_DURATION", "WAKE_AFTER_SLEEP_ON", "SLEEPLATENCY",
"REM_LATENCY", "SLEEPLATENCY_DEEPSLEEP")
# Helper to handle duplicate columns
dedup_cols <- function(df) {
colnames(df) <- make.unique(colnames(df))
df
}
# Custom loader for Adaptation files
read_adapt_only <- function(f) {
df <- readxl::read_excel(f, skip = 1, col_types = "text") %>%
dedup_cols()
df$Real_ID <- stringr::str_extract(basename(f), "^\\d{3}")
df$day <- "AD"
return(df)
}
# --- 2. EXECUTE DATA PROCESSING ---
library(tidyverse)
library(lubridate)
library(here)
# Load AD files
adapt_path <- here::here("Data", "raw", "EEG", "adaptation_night_sleep_profile")
adapt_files <- list.files(adapt_path, pattern = "\\.xlsx$", full.names = TRUE)
adapt_raw_df <- purrr::map_dfr(adapt_files, read_adapt_only)
## 1. Improved Converter: Handles Excel Serial (0.05), Seconds (4000), and Minutes (70)
convert_to_minutes_smart <- function(x) {
x_num <- as.numeric(as.character(x))
dplyr::case_when(
is.na(x_num) ~ NA_real_,
# If it's a tiny decimal (Excel serial), convert to minutes
x_num > 0 & x_num < 2 ~ x_num * 1440,
# If it's a massive number (Seconds), convert to minutes
x_num > 1440 ~ x_num / 60,
# Otherwise, assume it's already in minutes
TRUE ~ x_num
)
}
# 2. Clean the AD side
adapt_prep <- adapt_raw_df %>%
dplyr::rename(dplyr::any_of(rename_map)) %>%
# Apply the smart converter to every column except ID/Day
dplyr::mutate(across(-c(Real_ID, day), convert_to_minutes_smart)) %>%
dplyr::mutate(Real_ID = as.character(Real_ID), day = "AD")
# 3. Clean the BSL/Study side
# This forces 'clean' to be numeric too, stopping the "Can't combine" error
clean_prep <- clean %>%
dplyr::mutate(across(where(is.character) & !c(Real_ID, day), ~as.numeric(as.character(.x)))) %>%
dplyr::mutate(Real_ID = as.character(Real_ID), day = as.character(day))
# 4. Bind them - Now they are both numeric and in the right units!
full_timeline_df <- dplyr::bind_rows(adapt_prep, clean_prep) %>%
dplyr::mutate(day = factor(day, levels = c("AD", "BSL", "1", "2", "3", "4", "5", "6", "7", "FU")))
# --- 3. PIVOT & PLOT ---
full_long <- full_timeline_df %>%
dplyr::select(Real_ID, day, any_of(mins_plus_se)) %>%
# Force all metric columns to numeric
dplyr::mutate(across(-c(Real_ID, day), ~as.numeric(as.character(.x)))) %>%
tidyr::pivot_longer(
cols = -c(Real_ID, day),
names_to = "variable",
values_to = "value"
) %>%
dplyr::filter(!is.na(value))
full_summary <- full_long %>%
group_by(day, variable) %>%
summarise(
mean = mean(value, na.rm = TRUE),
se = sd(value, na.rm = TRUE) / sqrt(pmax(n(), 1)),
.groups = "drop"
)
p_full_timeline <- ggplot(full_summary, aes(x = day, y = mean, group = 1)) +
geom_line(linewidth = 0.8, color = "gray60", linetype = "dashed") +
geom_errorbar(aes(ymin = mean - se, ymax = mean + se), width = 0.2, color = "#2E4053") +
geom_point(size = 3, aes(color = (day == "AD")), shape = 16) +
scale_color_manual(values = c("TRUE" = "#E67E22", "FALSE" = "#2E4053")) +
facet_wrap(~ variable, scales = "free", ncol = 3) +
theme_classic() +
theme(
strip.background = element_blank(),
strip.text = element_text(face = "bold", size = 10),
axis.text.x = element_text(angle = 45, hjust = 1),
legend.position = "none"
) +
labs(x = "Study Night", y = "Mean ± SE")
## 1. Verify where 'here' thinks the root is
print(paste("Project Root is:", here::here()))
# 2. Save using the explicit path construction
# This ensures it goes: Project Root -> Figures -> EEG -> Filename
ggsave(
filename = here::here("Figures", "EEG", "EEG_with_adaptation_night.png"),
plot = p_full_timeline,
width = 15,
height = 9,
dpi = 300
)
```
# SOREMPS
```{r}
# 1. PREPARE THE DATA
soremp_clean <- full_timeline_df %>%
select(Real_ID, day, `REM Latency (mins)`) %>%
mutate(
rem_val = as.numeric(`REM Latency (mins)`),
# Define SOREMP: REM within 15 minutes of sleep onset
is_soremp = rem_val <= 15,
soremp_label = if_else(is_soremp, "SOREMP (≤15m)", "Normal (>15m)")
) %>%
# Filter out any NAs just in case, though you mentioned they all have REM
filter(!is.na(rem_val))
# 2. PLOT BY PARTICIPANT (To see individual drivers)
ggplot(soremp_clean, aes(x = day, y = 1, fill = soremp_label)) +
geom_tile(color = "white", linewidth = 0.5) + # Heatmap style tiles
facet_wrap(~ Real_ID, ncol = 3) + # One small plot per participant
scale_fill_manual(values = c("SOREMP (≤15m)" = "#d95f0e", "Normal (>15m)" = "#2E4053")) +
theme_minimal() +
labs(
title = "Individual SOREMP Occurrences across Study Nights",
subtitle = "Orange tiles indicate REM onset within 15 minutes",
x = "Study Night",
y = NULL,
fill = "Status"
) +
theme(
axis.text.y = element_blank(), # Hide Y axis since it's just a presence tile
axis.ticks.y = element_blank(),
panel.grid = element_blank(),
strip.text = element_text(face = "bold"),
axis.text.x = element_text(angle = 45, hjust = 1),
legend.position = "bottom"
)
# 3. SAVE INDIVIDUAL TRACKER
ggsave(
filename = here::here("Figures", "EEG", "SOREMP_Individual_Tracker.png"),
width = 12,
height = 8,
dpi = 300
)
# 1. CREATE ANONYMOUS MAPPING
# This creates a lookup table: Real_ID "101" -> "Participant A"
id_mapping <- full_timeline_df %>%
distinct(Real_ID) %>%
arrange(Real_ID) %>%
mutate(
anon_id = paste("Participant", LETTERS[row_number()])
)
# 2. PREPARE DATA WITH ANONYMOUS LABELS
soremp_anon <- full_timeline_df %>%
left_join(id_mapping, by = "Real_ID") %>%
select(anon_id, day, `REM Latency (mins)`) %>%
mutate(
rem_val = as.numeric(`REM Latency (mins)`),
is_soremp = rem_val <= 15,
soremp_label = if_else(is_soremp, "SOREMP (≤15m)", "REM Latency (>15m)")
) %>%
filter(!is.na(rem_val))
# 3. PLOT ANONYMIZED FACETS
ggplot(soremp_anon, aes(x = day, y = 1, fill = soremp_label)) +
geom_tile(color = "white", linewidth = 0.6) +
facet_wrap(~ anon_id, ncol = 4) + # Grouped by letter A, B, C...
scale_fill_manual(
values = c("SOREMP (≤15m)" = "#d95f0e", "REM Latency (>15m)" = "#2E4053"),
name = "Status"
) +
theme_minimal() +
labs(
x = "Study Night",
y = NULL
) +
theme(
axis.text.y = element_blank(),
axis.ticks.y = element_blank(),
panel.grid = element_blank(),
strip.text = element_text(face = "bold", size = 10),
axis.text.x = element_text(angle = 45, hjust = 1),
legend.position = "bottom",
strip.background = element_rect(fill = "#f0f0f0", color = NA)
)
# 4. SAVE
ggsave(
filename = here::here("Figures", "EEG", "SOREMs.png"),
width = 10,
height = 6,
dpi = 300
)
```
# Session info
```{r session-info}
sessioninfo::session_info()
```

06_EEG.Rmd at commit e76059b, under CC-BY-4.0 · at the source

Overview

Authors: Emily C. Stanyer¹, Maëlwenn Le Roux¹, Rachel Sharman¹, Sofia I. Ribeiro Pereira¹, Shaun Davidson², Lionel Tarassenko², Colin A. Espie¹, Simon D. Kyle¹

ORCID iDs: Emily C. Stanyer, Maëlwenn Le Roux, Rachel Sharman, Sofia I. Ribeiro Pereira, Shaun Davidson, Lionel Tarassenko, Colin A. Espie, Simon D. Kyle

Institutions: University of Oxford (United Kingdom)

Journal: medRxiv (preprint)

Dates: published online 15 May 2026

Type: Preprint

License: CC BY

Identifiers: DOI 10.64898/2026.05.08.26348885 · OpenAlex W7161289311

Open access: green, a free copy (OpenAlex)

Status: code verified

Categories: EEG (modality), other (modality), depression (population), sleep disorders (population), clinical / translational (subfield)

Methods: Spectral & time-frequency, Statistics, Preprocessing, Evoked potentials, Single-unit activity, calcium imaging, Physiology & signal measures, Connectivity, Machine learning

Keywords: depression, mental health, EEG, wearables, CBT-I, sleep disorders

Topic: Sleep and related disorders (Experimental and Cognitive Psychology, Psychology), according to OpenAlex

Funding: Wellcome Trust (227684/Z/22/Z, 227093/Z/23/Z); National Institute for Health Research (NIHR) (NIHR203316, NIHR203667)

Citations: not cited yet (Europe PMC); 58 references in the paper

Abstract

Objectives: Self-applied, low-density EEG offers opportunities to examine sleep in the home environment, yet its feasibility during behavioural sleep interventions remains unexplored. This pilot study aimed to evaluate the feasibility and acceptability of a self-applied, low-density EEG device during sleep restriction therapy (SRT) and explore effects on sleep and affect.

Methods: Seventeen adults with insomnia and depressive symptoms completed a 2-week baseline and 4 weeks of SRT. The primary outcome was the proportion of expected EEG recordings completed and scoreable. Secondary outcomes included clinical measures, sleep continuity (sleep diary, actigraphy), sleep architecture (low-density EEG for 9 nights), power spectral density, and affect. Data were analysed with linear mixed models. Cohen’s d and 95% confidence intervals were reported.

Results: Feasibility was demonstrated (92% of expected EEG nights completed). SRT was associated with reductions in insomnia severity, depressive symptoms, negative affect, and increases in positive affect. Robust improvements were observed across treatment in sleep continuity (SOL, WASO, SE) from diary, which were paralleled by actigraphy. EEG revealed reduced TIB, TST, N1, N2, REM sleep, and REM latency during week one. Reductions in EEG-derived TIB and N1 sleep were maintained at night 28. There were no reliable differences for spectral or spindle measures.

Conclusions: These findings suggest that self-applied, low-density EEG during SRT is feasible, acceptable, and may capture sleep changes during treatment. They highlight the potential for multi-night monitoring of sleep interventions at home and elucidating mechanisms underlying therapeutic change.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 7 matches between paragraphs and lines of code.

emstanyer/The-RESTORE-Study

License: CC-BY-4.0

State: the link answers, verified on 28 September 2026

Evidence: files inventoried

Commit: e76059bfe5b8d8e7bab346c28f7a305a244f8171, 22 June 2026

Languages: R (12), MATLAB (2)

Size: 44 files, 14 scripts

Software Heritage: not archived

Found in: the end of the paper

Holds: README, license file, environment (renv.lock), 12 notebooks

Not found: CITATION.cff, tests, continuous integration, documentation

Tools: tidyverse (12 files), emmeans (6 files), ggplot2 (5 files), lme4 (5 files), patchwork (4 files), FieldTrip (2 files), ggpubr (2 files), Signal Processing Toolbox (2 files), Statistics and Machine Learning Toolbox (2 files), rstatix (2 files), broom (1 file), easystats (1 file), lmerTest (1 file)

Availability: 1 check, the latest on 28 September 2026: the link answers

28 September 2026: the link answers

16 files

Scripts/01_demographics.Rmd, R, 173 lines
Scripts/02_follow_up.Rmd, R, 196 lines, 1 match
Scripts/03_feasibility.Rmd, R, 180 lines
Scripts/04_PANAS.Rmd, R, 483 lines
Scripts/05_sleep_diary_actigraphy.Rmd, R, 702 lines, 1 match
Scripts/06_EEG.Rmd, R, 1,171 lines, 2 matches
Scripts/07_power_spectral_analysis.Rmd, R, 389 lines
Scripts/08_spindles_and_SOs.Rmd, R, 324 lines, 1 match
Scripts/09_adherence.Rmd, R, 524 lines
Scripts/10_NPCRA_light_exposure.Rmd, R, 2,032 lines
Scripts/11_hypnogram_trim.Rmd, R, 147 lines
Scripts/12_hypnogram_renaming.Rmd, R, 63 lines
Scripts/RESTORE_Spectral_Analysis_NREM.m, MATLAB, 271 lines, 1 match
Scripts/RESTORE_Spectral_Analysis_REM.m, MATLAB, 265 lines, 1 match
LICENSE, License, 6 lines
README.md, Text, 35 lines

At the source: github.com/emstanyer/The-RESTORE-Study

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
14 scripts, each with its path and the digest of its content;
7 matches between paragraphs of the paper and lines of the code (method lexical-v1);
neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data availability statement

The data that support the findings of this study will be made available on reasonable request from the authors.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 28 September 2026: the first record

Recorded: type, journal, dates, 8 authors, 6 keywords, 2 funders, 52 references.

Cite

This paper

Stanyer, E. C., Le Roux, M., Sharman, R., Pereira, S. I. R., Davidson, S., Tarassenko, L., Espie, C. A., & Kyle, S. D. (2026). Insights from nine nights of self-applied, low-density sleep EEG during sleep restriction therapy: a proof-of-concept evaluation. medRxiv (preprint). https://doi.org/10.64898/2026.05.08.26348885

BibTeX

@article{stanyer2026insights, author = {Stanyer, Emily C. and Le Roux, Maëlwenn and Sharman, Rachel and Pereira, Sofia I. Ribeiro and Davidson, Shaun and Tarassenko, Lionel and Espie, Colin A. and Kyle, Simon D.}, title = {{Insights from nine nights of self-applied, low-density sleep EEG during sleep restriction therapy: a proof-of-concept evaluation}}, journal = {medRxiv (preprint)}, year = {2026}, month = may, publisher = {medRxiv}, doi = {10.64898/2026.05.08.26348885}, url = {https://doi.org/10.64898/2026.05.08.26348885} }

RIS

TY - JOUR AU - Stanyer, Emily C. AU - Le Roux, Maëlwenn AU - Sharman, Rachel AU - Pereira, Sofia I. Ribeiro AU - Davidson, Shaun AU - Tarassenko, Lionel AU - Espie, Colin A. AU - Kyle, Simon D. TI - Insights from nine nights of self-applied, low-density sleep EEG during sleep restriction therapy: a proof-of-concept evaluation T2 - medRxiv (preprint) J2 - medRxiv PY - 2026 DA - 2026/05/15 PB - medRxiv DO - 10.64898/2026.05.08.26348885 UR - https://doi.org/10.64898/2026.05.08.26348885 ER -

CSL-JSON

{ "id": "10.64898/2026.05.08.26348885", "type": "article", "title": "Insights from nine nights of self-applied, low-density sleep EEG during sleep restriction therapy: a proof-of-concept evaluation", "container-title": "medRxiv (preprint)", "author": [ { "family": "Stanyer", "given": "Emily C." }, { "family": "Le Roux", "given": "Maëlwenn" }, { "family": "Sharman", "given": "Rachel" }, { "family": "Pereira", "given": "Sofia I. Ribeiro" }, { "family": "Davidson", "given": "Shaun" }, { "family": "Tarassenko", "given": "Lionel" }, { "family": "Espie", "given": "Colin A." }, { "family": "Kyle", "given": "Simon D." } ], "container-title-short": "medRxiv", "DOI": "10.64898/2026.05.08.26348885", "publisher": "medRxiv", "URL": "https://doi.org/10.64898/2026.05.08.26348885", "issued": { "date-parts": [ [ 2026, 5, 15 ] ] } }

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.