Ambulatory ECG noise reduction algorithm for conditional diffusion model based on multi-kernel convolutional...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Signal Processing

Background:

• Ambulatory electrocardiogram (ECG) testing is vital for cardiovascular disease management.
• Exercise ECG signals are prone to noise, hindering accurate analysis and potentially causing misdiagnosis.
• Common noise sources include electrode motion, baseline wander, and muscle artifacts.

Purpose of the Study:

• To develop and validate a novel deep learning-based method for effective noise reduction in ambulatory ECG signals.
• To improve the clarity and reliability of ECG waveforms for better clinical interpretation.

Main Methods:

• A novel deep learning approach utilizing an enhanced diffusion model network.
• Incorporation of conditional noise and a multi-kernel convolutional transformer network for noise prediction.
• Integration of the diffusion model's inverse process for noise suppression.

Main Results:

• The proposed method demonstrated superior noise reduction performance compared to eight state-of-the-art algorithms on the QT and MIT-BIH Noise Stress Test databases.
• Achieved optimal results in statistical, distance-based, and waveform visualization metrics.
• Showcased stable performance against electrode motion, baseline wander, muscle artifacts, and their combinations.

Conclusions:

• The novel deep learning method offers a significant advancement in reducing noise in ambulatory ECG signals.
• This technique holds promise for future applications in clinical dynamic ECG signal analysis, enhancing diagnostic capabilities.
• Improved ECG signal quality can lead to more accurate diagnoses and better patient outcomes for cardiovascular diseases.