Atrial Fibrillation (AF) Analysis & HRV Pipeline

Utilizing the MIT-BIH Atrial Fibrillation Database (AFDB)

This repository contains a robust biomedical signal processing pipeline designed to detect and analyze Atrial Fibrillation (AF) using ECG signals. The workflow covers the entire lifecycle of a medical data project, from raw data acquisition to Heart Rate Variability (HRV) feature extraction and statistical visualization.

Introduction

Atrial Fibrillation (AF) is characterized by rapid and irregular beating of the atrial chambers. This project aims to quantify these irregularities by analyzing ECG records specifically 04015, 04043, and 07879 from the PhysioNet AFDB. By comparing Normal Sinus Rhythm (NSR) against AF episodes, we can visualize and measure the physiological impact on heart rate stability.

The Pipeline

The project is organized into a 10-step modular Python pipeline located in the /src directory:

1. Data Acquisition: Automated downloading of .dat, .hea, and .atr files via WFDB.
2. Preprocessing: Applied baseline wander removal and noise reduction to clean the raw ECG signals.
3. R-Peak Detection: Implemented QRS complex identification using the NeuroKit2 library.
4. R-Peak Validation: Cross-referenced detected peaks against reference annotations for Quality Control (QC).
5. Rhythm Labeling: Segmented and labeled RR intervals as "Normal" or "AF" based on physician-verified database annotations.
6. Feature Extraction: Computed HRV metrics and heart rate statistics.
7. Visualization: Generated comparative plots and Poincaré maps to highlight rhythm differences.

Technical Implementation

• Signal Processing: Used scipy.signal for digital filtering and signal conditioning.
• HRV Analysis: Calculated key time-domain metrics:
  • Mean RR: Average time between consecutive R-peaks.
  • SDNN: Standard deviation of the NN intervals (reflects overall variability).
  • RMSSD: Root mean square of successive differences (reflects parasympathetic activity).
• Non-Linear Analysis: Created Poincaré Plots ($RR_{n}$ vs $RR_{n+1}$) to visualize the "predictability" of the heart's rhythm.

Results & Observations

The analysis revealed clear physiological markers for Atrial Fibrillation:

• Variability: AF segments showed significantly higher SDNN and RMSSD values compared to normal rhythm, indicating high irregularity.
• Heart Rate: AF episodes were characterized by shorter RR intervals and a noticeably higher average Heart Rate (HR).
• Poincaré Patterns:
  • Normal Rhythm: Displays a tight, cigar-shaped cluster along the identity line.
  • AF Rhythm: Displays a scattered "cloud" or "fan" distribution, representing the stochastic nature of the irregular intervals.

Project Structure

AFib Analysis/
├── src/            # 1-10 Step-by-step Python scripts
├── utils/          # Visualization scripts
├── Results/        # Generated plots
└── requirements.txt # Project dependencies