cgmguru: Complete CGM Analysis Workflow

In cgmguru: Advanced Continuous Glucose Monitoring Analysis with High-Performance C++ Backend

knitr::opts_chunk$set(collapse = TRUE, comment = "#>", fig.width = 10, fig.height = 6)
library(cgmguru)
library(iglu)
library(ggplot2)
library(dplyr)
set.seed(123)

cgmguru: Complete CGM Analysis Workflow

The cgmguru package provides comprehensive tools for analyzing Continuous Glucose Monitoring (CGM) data using the GRID (Glucose Rate Increase Detector) algorithm and related methodologies. This vignette demonstrates the complete workflow from basic analysis to advanced event detection.

Core functions at a glance

• grid: Implements GRID to detect rapid glucose rate increases (meal starts); returns grid vector and episode summaries.
• mod_grid: Re-applies GRID logic with a defined gap and window to produce a modified grid for maxima mapping.
• maxima_grid: Combines GRID and local maxima to identify postprandial peaks within a time window and summarizes episodes.
• detect_hyperglycemic_events: Finds Level 1/Level 2/Extended hyperglycemia episodes using consensus thresholds and durations.
• detect_hypoglycemic_events: Finds Level 1/Level 2/Extended hypoglycemia episodes and duration below 54 mg/dL where applicable.
• detect_all_events: Unified interface summarizing hypo/hyper event types, counts, and durations by subject.
• find_local_maxima: Detects peaks where glucose rises for two points before and falls for two points after the peak.
• find_max_after_hours / find_max_before_hours: Finds maximum glucose within a time window after/before given start points; returns indices and episode counts.
• find_min_after_hours / find_min_before_hours: Finds minimum glucose within a time window after/before given start points.
• find_new_maxima: Refines maxima around candidate grid points to improve peak identification.
• detect_between_maxima: Characterizes events occurring between identified maxima, including time-to-peak.
• excursion: Calculates excursions (>70 mg/dL rise within 2 hours, not preceded by <70 mg/dL), with counts and starts.
• start_finder: Locates R-based positions of episode starts in a 0/1 vector (1 after 0 or at start).
• transform_df: Maps GRID episode starts to maxima (within 4 hours) and merges results for downstream analysis.
• orderfast: Efficiently orders data by id and time for large CGM datasets.

Tip: See individual help pages for details and examples, for instance:

?grid
?detect_all_events

Loading Sample Data

We'll use two datasets from the iglu package to demonstrate different analysis scenarios:

# Load example datasets
data(example_data_5_subject)  # 5 subjects, 13,866 readings
data(example_data_hall)       # 19 subjects, 34,890 readings

# Display basic information about the datasets
cat("Dataset 1 (example_data_5_subject):\n")
cat("  Rows:", nrow(example_data_5_subject), "\n")
cat("  Subjects:", length(unique(example_data_5_subject$id)), "\n")
cat("  Time range:", as.character(range(example_data_5_subject$time)), "\n")
cat("  Glucose range:", range(example_data_5_subject$gl), "mg/dL\n\n")

cat("Dataset 2 (example_data_hall):\n")
cat("  Rows:", nrow(example_data_hall), "\n")
cat("  Subjects:", length(unique(example_data_hall$id)), "\n")
cat("  Time range:", as.character(range(example_data_hall$time)), "\n")
cat("  Glucose range:", range(example_data_hall$gl), "mg/dL\n")

# Show first few rows
head(example_data_5_subject)

1. Basic GRID Analysis

The GRID algorithm detects rapid glucose rate increases, which are often associated with meal consumption.

# Perform GRID analysis on the smaller dataset
grid_result <- grid(example_data_5_subject, gap = 15, threshold = 130)

# Display results
cat("GRID Analysis Results:\n")
cat("  Detected grid points:", nrow(grid_result$grid_vector), "\n")
cat("  Episode counts:\n")
print(grid_result$episode_counts)

# Show first few detected grid points
cat("\nFirst few detected grid points:\n")
head(grid_result$grid_vector)

2. Hyperglycemic Event Detection

Detect different levels of hyperglycemic events according to clinical guidelines.

# Level 1 Hyperglycemic events (≥15 consecutive minutes >180 mg/dL)
hyper_lv1 <- detect_hyperglycemic_events(
  example_data_5_subject, 
  start_gl = 180, 
  dur_length = 15, 
  end_length = 15, 
  end_gl = 180
)

# Level 2 Hyperglycemic events (≥15 consecutive minutes >250 mg/dL)
hyper_lv2 <- detect_hyperglycemic_events(
  example_data_5_subject, 
  start_gl = 250, 
  dur_length = 15, 
  end_length = 15, 
  end_gl = 250
)

# Extended Hyperglycemic events (default parameters)
hyper_extended <- detect_hyperglycemic_events(example_data_5_subject)

cat("Hyperglycemic Event Detection Results:\n")
cat("Level 1 Events (>180 mg/dL):\n")
print(hyper_lv1$events_total)

cat("\nLevel 2 Events (>250 mg/dL):\n")
print(hyper_lv2$events_total)

cat("\nExtended Events (default):\n")
print(hyper_extended$events_total)

# Show detailed events for first subject
cat("\nDetailed Level 1 Events for Subject", hyper_lv1$events_detailed$id[1], ":\n")
head(hyper_lv1$events_detailed[hyper_lv1$events_detailed$id == hyper_lv1$events_detailed$id[1], ])

3. Hypoglycemic Event Detection

Detect hypoglycemic events using different thresholds.

# Level 1 Hypoglycemic events (≤70 mg/dL)
hypo_lv1 <- detect_hypoglycemic_events(
  example_data_5_subject, 
  start_gl = 70, 
  dur_length = 15, 
  end_length = 15
)

# Level 2 Hypoglycemic events (≤54 mg/dL)
hypo_lv2 <- detect_hypoglycemic_events(
  example_data_5_subject, 
  start_gl = 54, 
  dur_length = 15, 
  end_length = 15
)

cat("Hypoglycemic Event Detection Results:\n")
cat("Level 1 Events (≤70 mg/dL):\n")
print(hypo_lv1$events_total)

cat("\nLevel 2 Events (≤54 mg/dL):\n")
print(hypo_lv2$events_total)

4. Comprehensive Event Detection

Detect all types of glycemic events in one analysis.

# Detect all events with 5-minute reading intervals
all_events <- detect_all_events(example_data_5_subject, reading_minutes = 5)

cat("Comprehensive Event Detection Results:\n")
print(all_events)

5. Local Maxima Detection

Identify local maxima in glucose time series, which are important for postprandial peak analysis.

# Find local maxima
maxima_result <- find_local_maxima(example_data_5_subject)

cat("Local Maxima Detection Results:\n")
cat("  Total local maxima found:", nrow(maxima_result$local_maxima_vector), "\n")
cat("  Merged results:", nrow(maxima_result$merged_results), "\n")

# Show first few maxima
head(maxima_result$local_maxima_vector)

6. Maxima-GRID Combined Analysis

Combine maxima detection with GRID analysis for comprehensive postprandial peak detection.

# Combined maxima and GRID analysis
maxima_grid_result <- maxima_grid(
  example_data_5_subject, 
  threshold = 130, 
  gap = 60, 
  hours = 2
)

cat("Maxima-GRID Combined Analysis Results:\n")
cat("  Detected maxima:", nrow(maxima_grid_result$results), "\n")
cat("  Episode counts:\n")
print(maxima_grid_result$episode_counts)

# Show first few results
head(maxima_grid_result$results)

7. Excursion Analysis

Analyze glucose excursions above a threshold.

# Excursion analysis
excursion_result <- excursion(example_data_5_subject, gap = 15)

cat("Excursion Analysis Results:\n")
cat("  Excursion vector length:", length(excursion_result$excursion_vector), "\n")
cat("  Episode counts:\n")
print(excursion_result$episode_counts)

# Show episode start information
head(excursion_result$episode_start)

8. Advanced Pipeline: Complete Workflow

Demonstrate the complete analysis pipeline using the larger dataset for more comprehensive results. Note: This section may take longer to run on some machines.

# Use the larger dataset for comprehensive analysis
cat("Running complete analysis pipeline on example_data_hall...\n")

# Step 1: GRID analysis
cat("Step 1: GRID Analysis\n")
grid_pipeline <- grid(example_data_hall, gap = 15, threshold = 130)
cat("  Detected", nrow(grid_pipeline$grid_vector), "grid points\n")

# Step 2: Local maxima detection
cat("Step 2: Local Maxima Detection\n")
maxima_pipeline <- find_local_maxima(example_data_hall)
cat("  Found", nrow(maxima_pipeline$local_maxima_vector), "local maxima\n")

# Step 3: Modified GRID analysis
cat("Step 3: Modified GRID Analysis\n")
mod_grid_pipeline <- mod_grid(
  example_data_hall, 
  grid_pipeline$grid_vector, 
  hours = 2, 
  gap = 15
)
cat("  Modified grid points:", nrow(mod_grid_pipeline$mod_grid_vector), "\n")

# Step 4: Find maximum points after modified GRID points
cat("Step 4: Finding Maximum Points After GRID Points\n")
max_after_pipeline <- find_max_after_hours(
  example_data_hall,
  mod_grid_pipeline$mod_grid_vector,
  hours = 2
)
cat("  Maximum points found:", length(max_after_pipeline$max_indices), "\n")

# Step 5: Find new maxima
cat("Step 5: Finding New Maxima\n")
new_maxima_pipeline <- find_new_maxima(
  example_data_hall,
  max_after_pipeline$max_indices,
  maxima_pipeline$local_maxima_vector
)
cat("  New maxima identified:", nrow(new_maxima_pipeline), "\n")

# Step 6: Transform dataframes
cat("Step 6: Transforming Dataframes\n")
transformed_pipeline <- transform_df(
  grid_pipeline$episode_start, 
  new_maxima_pipeline
)
cat("  Transformed dataframe rows:", nrow(transformed_pipeline), "\n")

# Step 7: Detect between maxima
cat("Step 7: Detecting Between Maxima\n")
between_maxima_pipeline <- detect_between_maxima(
  example_data_hall, 
  transformed_pipeline
)
cat("  Between maxima analysis completed\n")

cat("\nComplete pipeline executed successfully!\n")

9. Time-Based Analysis Functions

Demonstrate functions that find maximum and minimum values within specific time windows.

# Create a subset for demonstration
subset_data <- example_data_5_subject[example_data_5_subject$id == unique(example_data_5_subject$id)[1], ][1:100, ]

# Create start points for time-based analysis
start_points <- subset_data[seq(1, nrow(subset_data), by = 20), ]

cat("Time-Based Analysis Functions:\n")

# Find maximum after 1 hour
max_after <- find_max_after_hours(subset_data, start_points, hours = 1)
cat("  Max after 1 hour:", length(max_after$max_indices), "points\n")

# Find maximum before 1 hour
max_before <- find_max_before_hours(subset_data, start_points, hours = 1)
cat("  Max before 1 hour:", length(max_before$max_indices), "points\n")

# Find minimum after 1 hour
min_after <- find_min_after_hours(subset_data, start_points, hours = 1)
cat("  Min after 1 hour:", length(min_after$min_indices), "points\n")

# Find minimum before 1 hour
min_before <- find_min_before_hours(subset_data, start_points, hours = 1)
cat("  Min before 1 hour:", length(min_before$min_indices), "points\n")

10. Data Ordering Utility

Demonstrate the fast dataframe ordering utility.

# Create sample data with mixed order
sample_data <- data.frame(
  id = c("b", "a", "c", "a", "b"),
  time = as.POSIXct(c("2023-01-01 10:00:00", "2023-01-01 09:00:00", 
                      "2023-01-01 11:00:00", "2023-01-01 08:00:00", 
                      "2023-01-01 12:00:00"), tz = "UTC"),
  gl = c(120, 100, 140, 90, 130)
)

cat("Original data (unordered):\n")
print(sample_data)

# Order the data
ordered_data <- orderfast(sample_data)

cat("\nOrdered data:\n")
print(ordered_data)

11. Visualization Examples

Create visualizations to better understand the analysis results.

# Select one subject for visualization
subject_id <- unique(example_data_5_subject$id)[1]
subject_data <- example_data_5_subject[example_data_5_subject$id == subject_id, ]

# Create a comprehensive plot
p1 <- ggplot(subject_data, aes(x = time, y = gl)) +
  geom_line(color = "blue", alpha = 0.7, size = 0.5) +
  geom_hline(yintercept = 180, color = "red", linetype = "dashed", alpha = 0.8) +
  geom_hline(yintercept = 250, color = "darkred", linetype = "dashed", alpha = 0.8) +
  geom_hline(yintercept = 70, color = "orange", linetype = "dashed", alpha = 0.8) +
  geom_hline(yintercept = 54, color = "darkorange", linetype = "dashed", alpha = 0.8) +
  labs(title = paste("CGM Data for Subject", subject_id, "with Clinical Thresholds"),
       subtitle = "Red lines: Hyperglycemia thresholds (180, 250 mg/dL)\nOrange lines: Hypoglycemia thresholds (70, 54 mg/dL)",
       x = "Time", 
       y = "Glucose (mg/dL)") +
  theme_minimal() +
  theme(plot.title = element_text(size = 14, face = "bold"),
        plot.subtitle = element_text(size = 10))

print(p1)

# Create a summary plot showing event counts across subjects
event_summary <- hyper_lv1$events_total
event_summary$subject <- paste("Subject", event_summary$id)

p2 <- ggplot(event_summary, aes(x = subject, y = total_events)) +
  geom_col(fill = "steelblue", alpha = 0.7) +
  geom_text(aes(label = total_events), vjust = -0.5) +
  labs(title = "Level 1 Hyperglycemic Events by Subject",
       x = "Subject",
       y = "Number of Events") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

print(p2)

12. Performance Comparison

Compare performance between datasets of different sizes.

# Function to measure execution time
measure_time <- function(expr) {
  start_time <- Sys.time()
  result <- eval(expr)
  end_time <- Sys.time()
  return(list(result = result, time = as.numeric(end_time - start_time, units = "secs")))
}

cat("Performance Comparison:\n")

# Test on smaller dataset
cat("Small dataset (5 subjects, 13,866 readings):\n")
small_time <- measure_time(grid(example_data_5_subject, gap = 15, threshold = 130))
cat("  GRID analysis time:", round(small_time$time, 3), "seconds\n")

small_maxima_time <- measure_time(find_local_maxima(example_data_5_subject))
cat("  Local maxima time:", round(small_maxima_time$time, 3), "seconds\n")

# Test on larger dataset
cat("\nLarge dataset (19 subjects, 34,890 readings):\n")
large_time <- measure_time(grid(example_data_hall, gap = 15, threshold = 130))
cat("  GRID analysis time:", round(large_time$time, 3), "seconds\n")

large_maxima_time <- measure_time(find_local_maxima(example_data_hall))
cat("  Local maxima time:", round(large_maxima_time$time, 3), "seconds\n")

# Calculate efficiency
efficiency_ratio <- (large_time$time / large_time$result$episode_counts$total_episodes) / 
                    (small_time$time / small_time$result$episode_counts$total_episodes)
cat("\nEfficiency ratio (large/small):", round(efficiency_ratio, 2))

Summary

This vignette demonstrates the comprehensive capabilities of the cgmguru package:

Core Analysis Functions:

• GRID Analysis: Detects rapid glucose rate increases
• Event Detection: Identifies hyperglycemic and hypoglycemic events
• Local Maxima: Finds peaks in glucose time series
• Excursion Analysis: Analyzes glucose excursions above thresholds

Advanced Pipeline Functions:

• Maxima-GRID Combination: Integrated analysis for postprandial peaks
• Time-Based Analysis: Finds maximum/minimum values within time windows
• Complete Workflow: End-to-end analysis pipeline
• Data Transformation: Prepares data for downstream analysis

Utility Functions:

• Fast Ordering: Efficient dataframe sorting
• Input Validation: Robust error handling and parameter validation

Key Features:

• High Performance: C++ implementation for speed
• Clinical Relevance: Based on established CGM analysis methodologies
• Comprehensive Coverage: Handles various glycemic event types
• Flexible Parameters: Customizable thresholds and time windows
• Robust Validation: Input validation and error handling

The package is designed for both research and clinical applications, providing reliable and efficient tools for CGM data analysis. For more detailed function documentation, see help(package = "cgmguru").

References

• Battelino, T., et al. "Continuous glucose monitoring and metrics for clinical trials: an international consensus statement." The Lancet Diabetes & Endocrinology 11.1 (2023): 42-57.
• Harvey, Rebecca A., et al. "Design of the glucose rate increase detector: a meal detection module for the health monitoring system." Journal of diabetes science and technology 8.2 (2014): 307-320.
• Chun, E., et al. "iglu: interpreting glucose data from continuous glucose monitors." R package version 3.0 (2023).

cgmguru documentation built on Nov. 6, 2025, 1:07 a.m.