HRV Comparison Across Meditation & Breathing Techniques 🧘‍♂️💓

📅 Project Overview

This project analyzes Heart Rate Variability (HRV) during various meditation and breathing states using the PhysioNet HRV dataset. We compare Chi meditation, Kundalini Yoga, sleep, and metronomic breathing based on time-domain HRV metrics and use machine learning models to distinguish between these states.

---

🎓 Learning Objectives

• Analyze HRV using time-domain features (Mean HR, SDNN, RMSSD)
• Compare meditation vs. sleep vs. breathing via statistical tests (t-test, ANOVA)
• Visualize physiological differences across states using box/violin plots
• Train ML classifiers (KNN, Random Forest) to classify states
• Interpret results and draw cardiovascular health insights

---

🧾 Dataset

Source: PhysioNet: Heart Rate Oscillations During Meditation

Groups

• Chi Meditation: C1–C8 (pre and med)
• Kundalini Yoga: Y1–Y4 (pre and med)
• Normal Sleep: N1–N11
• Ironman : I1-I9
• Metronomic Breathing: M1–M14

Columns

• Elapsed Time (seconds)
• Instantaneous Heart Rate (bpm)

---

🗺️ Data Preprocessing Overview

The following diagrams illustrate the preprocessing steps used:

📌 Data Loading Structure

📌 Preprocessing Summary Flow

🧼 Sample Preprocessing Code

🔹 Load Raw Heart Rate Data

# Define file paths
input_file = r"...\data\chi\C1.pre"
output_file = r"...\dataset\o_0_chi\C1_pre_meditation.csv"

# Load text data
df = pd.read_csv(input_file, sep="\s+", header=None, names=["Time", "HeartRate"])
df["Subject"] = "C1_pre"
df["Group"] = "Chi"
df["State"] = "Pre_Meditation"
df.to_csv(output_file, index=False)

🔹 Merge Pre/Post Meditation Data

file_pre = r"...\C1_pre_meditation.csv"
file_med = r"...\C1_meditation.csv"

# Merge files
pre = pd.read_csv(file_pre)
med = pd.read_csv(file_med)
c1_merged = pd.concat([pre, med], ignore_index=True)
c1_merged.to_csv("...\C1_merged_data.csv", index=False)

🔹 Extract R-R Intervals (from .qrs files)

import wfdb, numpy as np, pandas as pd

# Load QRS peaks
def load_qrs_peaks(path):
    annotation = wfdb.rdann(path, 'qrs')
    return annotation.sample

# Compute R-R intervals
r_peaks = load_qrs_peaks("...\C1pre")
time_stamps = r_peaks[1:] / 250
rr_intervals = np.diff(r_peaks) / 250

# Create DataFrame and save
rr_df = pd.DataFrame({"Time": time_stamps, "RR_Interval": rr_intervals})
rr_df["Subject"] = "C1_pre_rr"
rr_df["Group"] = "Chi Meditation"
rr_df["State"] = "Pre-Meditation"
rr_df.to_csv("...\RR_C1_pre.csv", index=False)

🔹 Combine All RR Intervals for Chi Group

import glob

rr_files = glob.glob(r"...\RR_C*_merged.csv")
df_all_rr = pd.concat([pd.read_csv(f) for f in rr_files], ignore_index=True)
df_all_rr.to_csv(r"...\RR_Chi_meditation_Combined.csv", index=False)

---

🔬 Statistical Testing Code (ANOVA + T-test)

from scipy.stats import ttest_rel, f_oneway, shapiro

# Paired T-Test (Chi pre vs. med)
t_stat, p_val = ttest_rel(df_chi_pre["RMSSD"], df_chi_med["RMSSD"])
print("T-test Chi RMSSD: t =", t_stat, "p =", p_val)

# ANOVA (across groups)
f_stat, p_val = f_oneway(df_chi["RMSSD"], df_yoga["RMSSD"], df_sleep["RMSSD"], df_metron["RMSSD"])
print("ANOVA RMSSD: F =", f_stat, "p =", p_val)

# Normality check
shapiro_stat, shapiro_p = shapiro(df_chi_pre["RMSSD"])
print("Shapiro-Wilk p =", shapiro_p)

---

🔄 Preprocessing Flowchart

📁 data/
├── 0_0_chi/
│   ├── RR_data/
│   │   └── RR_C1-C8_merged_data.csv, RR_Chi_meditation_Combined.csv
│   └── HRV_data/
│       └── C1-C8_merged_data.csv, HRV_Chi_Meditation_Combined.csv
├── 0_1_ironman/
│   ├── RR_data/
│   │   └── RR_Ironman_Combined.csv
│   └── HRV_data/
│       └── HRV_Ironman_Meditation_Combined.csv
├── 0_2_metron/
│   ├── RR_data/
│   │   └── RR_Metron_Combined.csv
│   └── HRV_data/
│       └── HRV_Metron_Combined.csv
├── 0_3_normal/
│   ├── RR_data/
│   │   └── RR_Normal_Combined.csv
│   └── HRV_data/
│       └── HRV_Normal_Combined.csv
├── 0_4_yoga/
│   ├── RR_data/
│   │   └── RR_Yoga_Combined.csv
│   └── HRV_data/
│       └── HRV_Yoga_Combined.csv

📁 Merged Files → 📄 o_1_all_hrv_rr_metric_combined.csv

📊 Statistical Computing:
├── 📄 Mean_HR
├── 📄 SDNN
├── 📄 RMSSD

📉 Statistical Tests:
├── 📄 o_3_results_normality_chi
├── 📄 o_4_results_normality_yoga
├── 📄 o_5_t_test_results_chi_yoga
├── 📄 o_2_anova_results_chi_yoga

---

🔬 Methodology

✅ 1. HRV Feature Extraction

Time-domain metrics used:

• Mean Heart Rate
• SDNN – Standard Deviation of NN intervals
• RMSSD – Root Mean Square of Successive Differences

✅ 2. Statistical Analysis

• Paired t-tests: Chi/Yoga (pre vs. med)
• ANOVA: Across meditation, sleep, and breathing
• Normality check via Shapiro-Wilk
• Visualizations: Boxplots plots

✅ 3. Machine Learning

• Model: KNN and Random Forest
• Features: Time-domain HRV metrics
• Evaluation: Accuracy, Confusion Matrix, F1-score

✅ 4. Tools

• Language: Python
• Libraries: pandas, numpy, scipy, matplotlib, seaborn, scikit-learn, wfdb
• Environment: Jupyter Lab (analysis + code), RStudio (report writing)

---

📈 Expected Outcomes

• Identify which meditation technique leads to highest HRV change
• Assess similarity of meditation-induced HRV to sleep or metronomic breathing
• Build classifier that can predict state (Chi, Yoga, Sleep, Breathing)
• Offer insights on HRV as an indicator of relaxation and cardiovascular health

---

📂 Folder Structure

hrv_meditation_analysis/
├── README.md
├── data/
│   ├── chi/  # C1_pre.txt, C1_med.txt ...
│   ├── yoga/  # Y1_pre.txt, Y1_med.txt ...
│   ├── sleep/  # N1.txt ...
│   ├── metron/  # M1.txt ...
├── merged/
│   └── o_1_all_hrv_rr_metric_combined.csv
├── stats/
│   ├── normality/
│   │   ├── o_3_results_normality_chi.csv
│   │   └── o_4_results_normality_yoga.csv
│   ├── t_tests/o_5_t_test_results_chi_yoga.csv
│   └── anova/o_2_anova_results_chi_yoga.csv
├── notebooks/
│   └── HRV_Comparison.ipynb
├── hrv_utils.py

---

🧠 Summary

Project: HRV Analysis of Meditation, Sleep, and Breathing

Goal: Quantify and compare HRV effects of Chi meditation, Kundalini Yoga, sleep, and metronomic breathing.

Highlights:

• Preprocessed HRV data using PhysioNet
• Computed time-domain HRV features
• Conducted statistical tests (paired t-test, ANOVA)
• Visualized HRV across states
• Trained ML classifiers to identify state

---

📝 References

1. Tiwari et al. (2021) – Heart Rate Variability Analysis
2. Kim et al. (2018) – Stress and HRV
3. Peng et al. (1999) – Heart Rate Oscillations During Meditation
4. Goldberger et al. (2000) – PhysioNet Reference

For full dataset access, visit: https://physionet.org/content/meditation/1.0.0/

---

Feel free to extend the project using wearable HRV devices, deep learning models (e.g., LSTM), or real-time biofeedback systems! 🧘‍♀️📊