Annotated real and synthetic datasets for non-invasive foetal electrocardiography post-processing benchmarking

Area of Science:

• Biomedical Engineering
• Signal Processing
• Maternal-Fetal Medicine

Background:

• Non-invasive fetal electrocardiography (fECG) is crucial for monitoring fetal well-being.
• Low signal-to-noise ratio (SNR) in fECG signals poses a significant challenge.
• Advanced post-processing is often necessary to enhance fECG quality.

Purpose of the Study:

• To present an annotated dataset of real and synthetic fECG signals.
• To support the development and benchmarking of fECG post-processing techniques.
• To facilitate research in fetal QRS detection and fECG extraction methods.

Main Methods:

• Acquisition of 21 dual-channel, 15s real fECG signals from 17 pregnant women (21-27 weeks gestation).
• Generation of 40 synthetic 10s fECG signals using the FECGSYN tool.
• Inclusion of raw recordings, expert cardiologist annotations of fetal R-peaks, and clean fECG signals for morphology preservation analysis.
• All signals sampled at 2048 Hz.

Main Results:

• A comprehensive dataset comprising real and synthetic fECG signals with detailed annotations was created.
• The dataset includes raw data for exploring alternative fECG extraction algorithms.
• Clean signals are provided for evaluating the performance of post-processing methods.

Conclusions:

• The presented dataset serves as a valuable benchmark for fECG post-processing techniques.
• It offers raw data for researchers developing novel fECG extraction and fetal QRS detection algorithms.
• This resource aims to advance the field of non-invasive fetal monitoring.