IBS_S3

Non-Invasive Fetal ECG Extraction using Independent Component Analysis (ICA)

Project Overview

This project focuses on the Biomedical Signal Analysis of fetal electrocardiograms (fECG). The primary objective is to non-invasively extract the fetal ECG signal from abdominal maternal recordings where maternal and fetal heart activities are mixed with noise.

Using Independent Component Analysis (ICA), this project isolates the fetal ECG component from multichannel abdominal signals and validates the extraction by comparing it against a direct fetal scalp electrode recording (ground truth).

---

Problem Statement & Objectives

Problem: Fetal ECG signals acquired from the maternal abdomen are significantly weaker (~10× smaller) than maternal ECG signals and often overlap in frequency, making extraction difficult due to noise and maternal dominance.

Objective:

1. To non-invasively isolate the fetal ECG using Independent Component Analysis (ICA).
2. To validate the extracted signal by comparing it with the direct fetal reference signal.
3. To compute the Heart Rate (BPM) and detect R-peaks.

---

Dataset

The project utilizes the Abdominal and Direct Fetal ECG Database (ADFECGDB) from PhysioNet.

• Records: 5 records (r01.edf to r05.edf).
• Duration: 5 minutes per record.
• Sampling Rate: 1000 Hz.
• Channels:
  • 4 Abdominal Channels: Mixed maternal and fetal signals.
  • 1 Direct Fetal Channel: Scalp electrode signal (Ground Truth).

---

Methodology

The project pipeline follows these distinct stages:

1. Data Loading

• EDF (European Data Format) files are loaded using the mne or pyEDFlib library.
• Channel information and sampling frequencies are extracted.

2. Signal Preprocessing

To ensure stability for the ICA algorithm, the raw signals undergo:

• Bandpass Filtering: Applied between 0.5 Hz – 45 Hz to remove baseline drift (low freq) and power line interference/muscle noise (high freq).
• Normalization: Signals are standardized to have zero mean and unit variance.

3. Independent Component Analysis (ICA)

• Concept: ICA is a blind source separation technique. It assumes the abdominal signal is a linear mixture of statistically independent sources (Maternal Heart, Fetal Heart, Noise).
• Process: The algorithm decomposes the 4 input abdominal channels into 4 independent components.

4. Component Identification & Validation

• Each independent component is compared with the Direct Fetal ECG (reference channel).
• Correlation Analysis: The component with the highest correlation coefficient relative to the direct fECG is identified as the extracted fetal signal.

5. Post-Processing

• R-Peak Detection: Locating the peaks of the QRS complex in the extracted signal.
• Heart Rate Calculation: Computing Beats Per Minute (BPM).

---