A 1.69$\mu$J Highly Robust Cardiac Arrhythmia Monitoring Processor With Triple-Adaptive QRS Detector and Medically...

Area of Science:

• Biomedical Engineering
• Cardiovascular Health Technology
• Signal Processing

Background:

• Electrocardiogram (ECG) signals from wearables are prone to noise, hindering accurate arrhythmia detection.
• Artifacts like motion, baseline wander, and muscle interference compromise signal integrity.
• Robust signal processing is essential for reliable cardiac health monitoring in real-world applications.

Purpose of the Study:

• To develop a highly robust cardiac health monitoring processor for accurate arrhythmia detection from noisy ECG signals.
• To enhance the reliability of wearable cardiac monitoring systems.
• To improve arrhythmia classification accuracy and computational efficiency.

Main Methods:

• Proposed a cascaded triple-adaptive QRS detector with event-driven sampling for accurate QRS complex identification in noisy signals.
• Developed a hybrid neural network (HNN) classifier integrating LSTM and ANN architectures with pathological feature fusion.
• Fabricated a prototype using 65-nm CMOS process for performance evaluation.

Main Results:

• Achieved low power consumption (2.53 µW total, 0.072 µW dynamic) and a compact 0.99mm² area.
• Demonstrated high R-peak detection sensitivity (>97.38%) and precision (>97.08%) on the MIT-BIH Noise Stress Test Database.
• Exceeded 90.1% inter-patient arrhythmia classification accuracy on the MIT-BIH Arrhythmia Database under low SNR conditions.

Conclusions:

• The developed processor offers a robust solution for arrhythmia detection in noisy ECG signals from wearable devices.
• The hybrid neural network approach provides efficient and accurate arrhythmia classification.
• The system's low power and computational complexity make it suitable for integration into wearable health monitoring devices.