PEPbench-Open, Reproducible, and Systematic Benchmarking of Automated Pre-Ejection Period Extraction Algorithms

Area of Science:

• Psychophysiology
• Biomedical Signal Processing
• Cardiovascular Physiology

Background:

• The pre-ejection period (PEP) is a key cardiac parameter in psychophysiology, often used to assess sympathetic nervous system (SNS) activity.
• Automated algorithms for PEP extraction from ECG and impedance cardiography (ICG) signals exist but lack systematic benchmarking.
• A significant barrier to progress is the absence of open-source algorithms and standardized, annotated datasets.

Purpose of the Study:

• To introduce PEPbench, an open-source Python package and evaluation framework for PEP extraction algorithms.
• To systematically benchmark various PEP extraction algorithm combinations using a standardized approach.
• To provide the first publicly available datasets with reference annotations for Q-peaks and B-points.

Main Methods:

• Developed PEPbench, integrating diverse Q-peak and B-point detection algorithms into comprehensive pipelines.
• Conducted a systematic comparison of 108 distinct algorithm combinations using PEPbench.
• Evaluated all combinations on two newly released, manually annotated datasets for Q-peaks and B-points.

Main Results:

• Algorithm performance varied significantly, with B-point detection contributing substantial error.
• Automated PEP extraction algorithms demonstrated relatively high error rates on a beat-to-beat basis.
• The study underscores the need for caution when applying automated PEP extraction in real-time psychophysiological research.

Conclusions:

• PEPbench provides a crucial open-source tool for reproducible benchmarking of PEP extraction algorithms.
• The findings highlight the limitations of current automated PEP extraction methods, particularly for beat-to-beat analysis.
• Encourages community contribution to foster innovation and improve the reliability of cardiac parameter extraction in psychophysiology.