Improved peaks automatic detection algorithm of noisy quasi-periodic physiological signals

Area of Science:

• Signal Processing
• Biomedical Engineering
• Computational Science

Background:

• The Automatic Multiscale-based Peak Detection (AMPD) method is computationally intensive and prone to errors.
• Existing methods struggle with noise and efficiency in quasi-periodic signal analysis.

Purpose of the Study:

• To develop an Improved AMPD (IAMPD) algorithm for faster and more robust peak detection in quasi-periodic signals.
• To enhance computational efficiency and maintain high accuracy, especially in noisy environments.

Main Methods:

• Implemented a frequency-informed scale constraint to reduce search space and redundant computations.
• Optimized the computational process by integrating operations and minimizing caching.
• Evaluated IAMPD on simulated signals with varying SNRs and real physiological data.

Main Results:

• IAMPD achieved a speedup of over 160 times compared to the original AMPD.
• Maintained 100% performance in Sensitivity (Se), Positive Predictivity (+P), and F1-score (F1) even at 0 dB SNR.
• Demonstrated superior performance over benchmark methods in peak detection accuracy and robustness.

Conclusions:

• IAMPD significantly improves computational speed and noise robustness for peak detection in quasi-periodic signals.
• The algorithm retains AMPD's parameter-free nature, requiring only an estimated frequency range.
• IAMPD is effective for real-time applications, accurately detecting peaks in various physiological signals like ECG and heart sounds.