Using the Redundant Convolutional Encoder-Decoder to Denoise QRS Complexes in ECG Signals Recorded with an Armband...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Long-term electrocardiogram (ECG) recordings are crucial for analyzing heart rate variability and detecting arrhythmias.
• Motion artifacts and noise frequently corrupt ECG signals from wearable devices, leading to inaccurate heart rate calculations.
• Accurate ECG signal processing is essential for reliable clinical interpretation.

Purpose of the Study:

• To develop and evaluate a denoising algorithm for ECG signals acquired using a wearable armband device.
• To assess the algorithm's effectiveness in removing motion artifacts and added noise.
• To improve the accuracy of R-peak detection and signal quality for long-term ECG monitoring.

Main Methods:

• A redundant convolutional encoder-decoder (R-CED) fully convolutional network was employed for denoising.
• ECG signals were recorded for 24 hours from a single participant using a wearable armband.
• Performance was evaluated by R-peak detection accuracy and signal quality indices (SNR, ratio of power, cross-correlation) on clean, noisy, and denoised sequences.
• The algorithm was also tested on the MIT-BIH arrhythmia database.

Main Results:

• The denoising algorithm significantly improved R-peak detection rates for both added noise (70-100% vs. 34-97%) and motion artifacts (91.86% vs. 61.16%).
• Signal-to-noise ratio (SNR) improved by 7-19 dB for signals with added noise and 0.39 dB for motion artifacts.
• The algorithm demonstrated a 7.08 ± 0.25 dB SNR improvement on the MIT-BIH arrhythmia database, indicating broad applicability.

Conclusions:

• The proposed R-CED algorithm effectively denoises ECG signals corrupted by motion artifacts and noise.
• This method enhances R-peak detection accuracy and overall signal quality, crucial for clinical applications.
• The algorithm shows potential for integration with various ECG measurement devices to improve diagnostic reliability.