Statistical model for cardiovascular signals with independent respiratory modulation for tracking pulse pressure variation

Cardiovascular signals including the electrocardiogram, pressure signals, and photoplethysmographs such as those used in pulse oximetry contain a wealth of information. Statistical models of these signals provide a means of representing and quantifying this information, and often lead to natural and optimal estimation algorithms. One powerful statistical model uses a Fourier approach to model cardiovascular signals as a harmonic sum of sinusoids with a fundamental frequency, amplitudes, and phases that vary slowly over time. We have further developed this model to incorporate respiratory effects including an additive component, pulse pressure variation (PPV), and respiratory sinus arrhythmia. PPV may be viewed as a form of amplitude modulation of the cardiovascular signal due to respiration. Current models do not explain the asymmetry between the upper and lower envelopes observed in cardiovascular pressure signals, and consequently are not appropriate for PPV estimation. We propose a model in which each of the cardiac harmonics is independently modulated by the respiratory signal. This improves the estimation accuracy and permits more accurate cardiovascular tracking and estimation. The proposed model is more accurate in PPV estimation applications.