A denoising method for ECG signals based on CEEMDAN-TSO and stacked sparse autoencoders

Area of Science:

• Biomedical Engineering
• Signal Processing

Background:

• Electrocardiogram (ECG) signals are vital for diagnosing heart conditions.
• Environmental and equipment factors introduce noise, obscuring critical diagnostic information in ECGs.
• Effective noise reduction is essential for accurate ECG analysis.

Purpose of the Study:

• To propose a novel noise-reduction method for low-frequency ECG signals.
• To enhance the accuracy of cardiovascular disease diagnosis through improved ECG signal quality.

Main Methods:

• Developed the CEEMDAN-TSO-SSAE method, integrating Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Tuna Swarm Optimization (TSO), and Stacked Sparse Autoencoder (SSAE).
• TSO optimized CEEMDAN parameters (Noise Standard Deviation, Number of Realizations, Maximum Iterations).
• Intrinsic Mode Functions (IMFs) were screened using the correlation coefficient method, and effective IMFs were denoised with SSAE before signal reconstruction.

Main Results:

• The CEEMDAN-TSO-SSAE method achieved the highest Signal-to-Noise Ratio (SNR) of 19.88 and the lowest Mean Squared Error (MSE) of 0.02 on simulated signals.
• On real ECG signals with baseline drift, the method yielded an SNR of 20.25 and an MSE of 0.01.
• Outperformed wavelet packet decomposition, Empirical Mode Decomposition, TSO-based Variational Mode Decomposition, and Denoising Autoencoder algorithms.

Conclusions:

• The CEEMDAN-TSO-SSAE method effectively eliminates complex noise from ECG signals.
• The proposed method preserves essential signal components crucial for accurate diagnosis.
• This technique offers a significant advancement in ECG signal processing for cardiovascular health monitoring.