An ECG Signal Denoising Method Using Conditional Generative Adversarial Net.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Electrocardiogram (ECG) signals are crucial for diagnosing cardiac conditions.
• Noise interference significantly degrades ECG signal quality and diagnostic accuracy.
• Existing denoising methods struggle with diverse and multiple noise types.

Purpose of the Study:

• To propose a novel denoising method for ECG signals using an improved conditional generative adversarial network (CGAN).
• To enhance the performance and availability of ECG denoising under various noise conditions.
• To validate the effectiveness of the proposed method in improving cardiac disease classification.

Main Methods:

• Developed a denoising framework based on conditional generative adversarial network (CGAN).
• Designed a generator using an optimized convolutional auto-encoder (CAE) to preserve signal features.
• Employed a discriminator with convolutional and fully connected layers for noise identification.

Main Results:

• Achieved an average signal-to-noise ratio (SNR) above 39 dB for denoised ECG signals across single and mixed noise scenarios.
• Demonstrated superior performance compared to state-of-the-art denoising methods.
• Significantly improved average accuracy for classifying four cardiac diseases by over 32% under multiple noises at SNR=0 dB.

Conclusions:

• The proposed CGAN-based method effectively removes noise from ECG signals while preserving essential diagnostic features.
• The method exhibits strong performance and generalization capabilities, outperforming existing techniques.
• This approach holds significant potential for improving the reliability of ECG-based diagnostics in noisy environments.