Pulse Wave Curve Fitting to Heterogeneous Noninvasive Plethysmographic Signals for Blood Pressure Tracking

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Wearable Technology

Background:

• Arterial blood pressure (BP) is a critical vital sign, increasingly relevant for wearable sensor applications.
• Signals from wearable sensors are often low-quality, necessitating tailored algorithms for surrogate BP parameters.
• Existing methods require robust parameter extraction for accurate, real-time BP estimation.

Purpose of the Study:

• To investigate unimodal pulse wave signals from diverse sources (photoplethysmography, bioimpedance, tonometry) for BP correlation.
• To derive and evaluate multiple parameters, including curve fitting features, against reference BP.
• To assess the performance of different signal processing techniques for wearable BP monitoring.

Main Methods:

• Acquisition of unimodal pulse waves from photoplethysmography (PPG), bioimpedance, and pulse applanation tonometry.
• Derivation and evaluation of multiple parameters, including slope transit time and curve fitting (Lognormal, Weibull, Gompertz) via linear least squares regression.
• Correlation analysis using Spearman Rho coefficients and assessment of mean absolute deviation and time lags against reference BP.

Main Results:

• Single parameters achieved Spearman Rho coefficients up to 0.78 (average 0.55) with significant p-values.
• Mean absolute deviation reached 0.08, with time lags between 0-30 seconds (mean 25 seconds).
• Curve fitting parameters demonstrated greater robustness than single-point parameters; PPG wave features exhibited superior BP tracking.

Conclusions:

• Curve fitting parameters derived from pulse wave analysis are more robust for BP estimation than single-point metrics.
• Photoplethysmography (PPG) wave features show the highest potential for accurate blood pressure tracking in wearable applications.
• Further investigation is needed to understand the inconsistency in correlation coefficient signs across parameters.