Using Kalman Filtering to Predict Time-Varying Parameters in a Model Predicting Baroreflex Regulation During Head-Up Tilt

Using Kalman Filtering to Predict Time-Varying Parameters in a Model Predicting Baroreflex Regulation During Head-Up Tilt 292 410 IEEE Transactions on Biomedical Engineering (TBME)

Using Kalman Filtering to Predict Time-Varying Parameters in a Model Predicting Baroreflex Regulation During Head-Up Tilt

The cardiovascular control system is constantly utilized to distribute and maintain required amounts of oxygenated blood to vitals areas in order to maintain homeostasis. However, in a subset of the population with orthostatic intolerance, this system can fail. The reasons for this failure are not well understood and determining which mechanisms are impaired is difficult. This problem is further challenging due to the variation observed within and between groups, sparse clinical data, typically only consisting of heart rate and blood pressure, and current assessments largely being static (e.g. patients with high blood pressure are hypertensive; do not account for how blood pressures changes in response to a stimulus) rather than dynamic. This study aims to address these challenges by building a physiological model that describes how relevant factors (heart rate, vascular resistance, cardiac contractility, etc) dynamically change and compare those dynamics between groups. More specifically, this study compares two methods for predicting time-varying changes in cardiac contractility and vascular resistance during head up tilt. The first method uses piecewise linear splines, similar to the method used by Williams et al [1], while the second method utilizes the ensemble Kalman filter. To quantify the uncertainty within these estimates, the delayed rejection metropolis hastings algorithm is implemented for the spline methodology. The estimates and uncertainties are compared for the spline and filter; while the spline is easier to implement, the filter is significantly faster computationally and includes uncertainties for the estimates.