Related Experiment Video
Updated: Dec 31, 2025
08:22
Author Spotlight: Advancing the Study of Brain-Heart Interplay with a Comprehensive EEGLAB Plugin for Multimodal Signal Analysis
Published on: April 26, 2024
2.9K
End-to-end trained encoder-decoder convolutional neural network for fetal electrocardiogram signal denoising
Eleni Fotiadou1, Tomasz Konopczyński2, Jürgen Hesser2
1Department of Electrical Engineering, Eindhoven University of Technology, Eindhoven 5612 AP, The Netherlands.
Physiological Measurement
|January 10, 2020
Summary
Artificial intelligence enhances fetal electrocardiograms (ECGs) by improving signal quality for better fetal health assessment. This deep learning approach significantly boosts signal-to-noise ratio, aiding clinical decisions.
Area of Science:
- Biomedical Engineering
- Artificial Intelligence
- Signal Processing
Background:
- Non-invasive fetal electrocardiography (ECG) offers crucial fetal health insights.
- Low signal-to-noise ratio (SNR) in fetal ECG hinders clinical application.
- Improving fetal ECG quality is vital for accurate medical decision-making.
Purpose of the Study:
- To enhance single-channel fetal ECG signals using artificial intelligence.
- To develop a post-processing technique following maternal ECG suppression.
- To improve the diagnostic value of non-invasive fetal ECG.
Main Methods:
- A deep, fully convolutional encoder-decoder neural network framework was developed.
- The network learns end-to-end mappings from noisy to clean fetal ECG signals.
- Symmetric skip-layer connections were utilized to preserve signal details.
Main Results:
- An average SNR improvement of 7.5 dB was achieved for synthetic data.
- Real-world signal analysis showed low root mean square errors for PR (9.9 ms) and QT (14 ms) intervals.
- Substantial noise reduction was demonstrated on both synthetic and real fetal ECG data.
Conclusions:
- The AI method effectively enhances fetal ECG quality, preserving pulse shape and beat-to-beat variations.
- No prior knowledge of noise spectra or pulse locations is required.
- The technique shows promise for improving non-invasive fetal monitoring and diagnostics.
