The Deep-Match Framework: R-Peak Detection in Ear-ECG.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Wearable Technology

Background:

• Ear-based electrocardiograms (Ear-ECG) offer enhanced comfort and wearability for continuous heart monitoring.
• Signal quality degradation in wearable ECG, including Ear-ECG, is a significant challenge for practical applications.
• Accurate R-peak detection is crucial for reliable ECG analysis and interpretation.

Purpose of the Study:

• To introduce and evaluate a Deep Matched Filter (Deep-MF) for accurate R-peak detection in noisy wearable ECG signals.
• To improve the real-world functionality and reliability of Ear-ECG technology.
• To enhance the acceptance of deep learning models in e-Health through interpretable methods.

Main Methods:

• Development of a Deep Matched Filter (Deep-MF) comprising an ECG template-initialized encoder and an R-peak classifier.
• Utilizing the encoder as a Matched Filter to identify ECG template matches within the input signal.
• Employing convolutional layers for filtering and a classifier for precise R-peak localization.

Main Results:

• The Deep-MF achieved a median R-peak recall of 94.9% and a median precision of 91.2% across 36 subjects using leave-one-subject-out cross-validation.
• The proposed method demonstrated superior performance compared to existing algorithms for R-peak detection in noisy wearable ECG.
• The Deep-MF effectively addresses signal quality degradation issues common in wearable health technologies.

Conclusions:

• The Deep Matched Filter framework significantly advances the practical utility of Ear-ECG systems.
• The interpretable nature of the Deep-MF promotes trust and adoption of deep learning in e-Health applications.
• This research paves the way for more reliable and user-friendly wearable cardiovascular monitoring solutions.