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

Arterial blood pressure is an important vital sign, and is becoming relevant for wearable sensors. Commonly, the signals recorded in this context are of poor quality and the algorithms working on surrogate parameters must be tailored thereto. In our current work we investigate several unimodal pulse waves acquired from three heterogeneous sources: photoplethysmography, bioimpedance and pulse applanation tonometry. We derive and evaluate multiple parameters regarding their correlation to reference blood pressure. One benchmark feature is the slope transit time. Parameters stem from fitting Lognormal, Weibull and Gompertz curves to the data using the linear least squares regression. Spearman Rho coefficients of up to 0.78 and averaging 0.55 at highly significant p-values are recorded for single parameters. The mean absolute deviation reaches 0.08. The results indicate there are 0 to 30 second lags between reference and parameter curves, usually with 25 seconds mean absolute deviations. The sign of the correlation coefficients is consistent only for a small subset of parameters, the underlying cause could not yet be identified. We conclude that the curve fitting parameters are more robust than single point ones, and PPG wave features perform best at blood pressure tracking.