A novel ECG QRS complex detection algorithm based on dynamic Bayesian network.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Accurate QRS complex detection is vital for electrocardiogram (ECG) analysis, but current wearable devices struggle with noise interference.
• Existing methods often focus solely on ECG waveforms, limiting their robustness against complex noise.

Purpose of the Study:

• To develop a novel QRS complex detection method for wearable ECG devices that enhances noise robustness and accuracy.
• To integrate ECG waveform and heart rhythm information into a unified probabilistic model.

Main Methods:

• A dynamic Bayesian network (DBN) approach was developed, incorporating the probability distribution of RR intervals.
• Unsupervised parameter optimization using expectation maximization (EM) was employed for patient-specific adaptation.
• Simplification strategies and an online detection mode were implemented for improved efficiency and real-time capability.

Main Results:

• The proposed DBN-based method demonstrated superior performance compared to state-of-the-art methods, including deep learning (DL) approaches, particularly on noisy datasets.
• The algorithm showed high accuracy, noise robustness, generalization ability, and real-time processing capabilities.

Conclusions:

• The DBN-based QRS detection algorithm offers a promising solution for accurate and robust heartbeat localization in wearable ECG devices.
• The method's accuracy, noise resilience, and scalability suggest significant potential for clinical applications in remote patient monitoring.