PPG Beat Segmentation Algorithms
PPG beat segmentation divides a continuous PPG signal into individual cardiac cycle waveforms by detecting beat boundaries at pulse onset (foot) points. Accurate segmentation is prerequisite for morphological analysis, enabling extraction of systolic time, diastolic time, pulse area, and waveform shape features.
Beat segmentation typically identifies pulse onset (foot) points as beat boundaries using one of several methods: first-derivative zero-crossing (the foot corresponds to the minimum of the first derivative before systolic upstroke), intersecting tangent method (intersection of the baseline tangent and the steepest upstroke tangent), or second-derivative maximum (the maximum acceleration point marking the start of rapid systolic filling).
The intersecting tangent method is preferred for morphological analysis because it provides consistent timing relative to arterial pressure waveform onset, with <5 ms variability across signal quality levels. First-derivative methods are faster computationally but more sensitive to baseline wander. For automated pipelines, a two-pass approach first detects systolic peaks (using Pan-Tompkins or template matching), then refines onset timing within a search window preceding each peak.
Quality-aware segmentation assigns a confidence score to each beat boundary and rejects beats with ambiguous boundaries. Beats with unclear onset points (common during baseline wander or motion artifacts) are excluded before downstream morphological analysis. Ensemble segmentation combining multiple onset detection methods and selecting the consensus boundary achieves onset timing accuracy of 3–8 ms relative to invasive arterial pressure waveform onset in ICU validation studies.
Frequently Asked Questions
What is the difference between peak-based and onset-based segmentation?
Peak-based segmentation uses systolic peaks as boundaries, while onset-based uses pulse foot points. Onset-based is preferred for morphological analysis because beat duration and timing features are referenced to the start of the cardiac cycle.
How does respiratory modulation affect beat segmentation?
Respiratory modulation causes slow amplitude and baseline variations that can shift apparent onset positions. High-pass filtering (>0.5 Hz) or baseline subtraction before onset detection minimizes respiratory effects on segmentation accuracy.
Can beat segmentation work during atrial fibrillation?
Yes, but with reduced accuracy. AF produces variable beat amplitudes and irregular timing, making fixed-window approaches unreliable. Adaptive methods that adjust search windows based on recent IBI history maintain >90% segmentation accuracy during AF.