Robust QRS detection based on simulated degenerate optical parametric oscillator-assisted neural network

Area of Science:

• Biomedical Engineering
• Cardiovascular Signal Processing
• Artificial Intelligence in Healthcare

Background:

• Accurate QRS complex detection in electrocardiography (ECG) is crucial for diagnosing cardiovascular diseases.
• Neural network approaches show promise for QRS detection but are sensitive to noise in ECG recordings.
• Existing methods struggle with performance degradation due to inevitable noise during ECG acquisition.

Purpose of the Study:

• To enhance the robustness and performance of neural network-based QRS detectors against noise.
• To introduce a novel Simulated Degeneration Unit (SDU)-assisted Convolutional Neural Network (CNN) for improved QRS detection.

Main Methods:

• Developed a CNN model incorporating a Simulated Degeneration Unit (SDU).
• The SDU simulates optical pulse degeneration to suppress in-band noise effectively.
• Evaluated the SDU-enhanced CNN on three open-source ECG databases and through real-world noise injection tests.

Main Results:

• The SDU-enhanced CNN demonstrated superior performance in QRS complex detection compared to other recent methods.
• Comprehensive evaluations on multiple databases confirmed the effectiveness of the proposed approach.
• Noise injection tests revealed a 167-300% higher optimal noise robustness boundary for the SDU-equipped CNN.

Conclusions:

• The SDU-assisted CNN offers a significant advancement in robust and accurate QRS detection.
• This method provides enhanced noise resilience, crucial for reliable ECG analysis in clinical settings.
• The SDU-enhanced CNN represents a promising tool for improving cardiovascular disease detection via ECG.