Pan-Tompkins Algorithm Adapted for PPG
The Pan-Tompkins algorithm, originally designed for ECG QRS detection, has been adapted for PPG systolic peak detection through modified bandpass filtering (0.5–8 Hz for PPG vs. 5–15 Hz for ECG), derivative-based slope enhancement, signal squaring, moving window integration, and adaptive dual-threshold decision logic.
The original Pan-Tompkins pipeline for ECG consists of five stages: bandpass filter (5–15 Hz), differentiation, squaring, moving window integration, and adaptive thresholding with search-back. For PPG adaptation, the bandpass filter is shifted to 0.5–8 Hz to capture the lower-frequency cardiac pulse waveform. The derivative stage enhances the steep systolic upstroke characteristic of PPG, and squaring emphasizes these high-slope regions while suppressing low-amplitude noise.
The adaptive thresholding mechanism maintains two thresholds — a signal threshold and a noise threshold — updated based on running peak and noise level estimates. Peaks exceeding the signal threshold are classified as systolic peaks. If no peak is detected within 166% of the running R-R interval estimate, a search-back mechanism lowers the threshold to recover missed beats, crucial for maintaining detection during transient artifacts.
PPG-specific modifications include an additional amplitude consistency check (rejecting peaks whose amplitude deviates >50% from the running average), a refractory period of 200 ms to prevent double-detection on the dicrotic notch, and slope-ratio validation comparing the systolic upstroke slope to the diastolic downstroke. These modifications achieve F1 scores of 0.96–0.99 on clean PPG and 0.88–0.94 on motion-corrupted signals in standard benchmarks including the MIMIC-III waveform database.
Frequently Asked Questions
Why adapt Pan-Tompkins for PPG instead of using PPG-specific detectors?
Pan-Tompkins provides a well-validated, computationally efficient framework. Its adaptive thresholding and search-back mechanisms are directly applicable to PPG. The adaptation requires only filter parameter changes and additional PPG-specific validation rules.
What is the detection latency of Pan-Tompkins on PPG?
The moving window integration introduces a delay equal to half the window width (typically 75–150 ms). Total detection latency is 100–200 ms from peak occurrence, acceptable for real-time heart rate display but relevant for precise IBI timing.
How does Pan-Tompkins handle PPG dicrotic notch false detections?
A refractory period of 200 ms after each detected peak prevents dicrotic notch detection. Additionally, slope-ratio validation ensures the upstroke-to-downstroke ratio matches systolic peak characteristics rather than dicrotic notch morphology.