Automatic Peak-to-Peak Detection and Segmentation of Radial Pulse Waveforms

Abstract: Accurate identification of systolic peaks and individual pulse cycles is important for reliable analysis of radial pulse waveforms. This study presents an automated peak-to-peak based method for systolic peak detection and pulse-cycle segmentation from real-time radial pulse signals. Radial pulse recordings were acquired using a Nellcor SpO₂ probe at a sampling frequency of 500 Hz, with three-minute real-time PPG recordings obtained from 25 samples. The acquired signals were subjected to baseline correction, smoothing, Savitzky–Golay filtering, and normalization to reduce signal variations while retaining relevant waveform characteristics. Systolic peaks were identified using a prominence-based detection approach incorporating signal-dependent amplitude characteristics and minimum peak- distance criteria. Consecutive detected peaks were used to determine peak-to-peak intervals and adaptive boundaries for pulse-cycle segmentation, followed by valley-based refinement of the segment boundaries. The peak-detection performance was quantitatively assessed using true-positive, false-positive, and false-negative detections together with sensitivity, precision, detection error rate, and F1-score. Additionally, 25 impedance plethysmography (IPG) samples representing five drug conditions were used to compare the peak-detection results obtained using the proposed method with those obtained using BARC software as the reference method. The proposed method provides an automated approach for consistent peak identification and segmentation of radial pulse waveforms.

Keywords: Radial Pulse Waveform, Peak-to-Peak Detection, Pulse Segmentation, Savitzky-Golay Filtering, Pulse Cycle Extraction, Real-time Pulse Analysis.