Point Process Models of Human Heart Beat Interval Dynamics


Supported by NIH Grant R01-HL084502

PRINCIPAL INVESTIGATOR

Riccardo Barbieri, Ph.D.

Abstract

Heart rate is a moment-to-moment indicator of cardiovascular integrity measured on every physical examination. Heart rate is also monitored continuously in patients under anesthesia, during surgery, in those treated in an intensive care unit and in fetuses during labor. Heart rate variability is an important quantitative marker of cardiovascular regulation by the autonomic nervous system that is widely used in research studies, as well as in clinical practice to diagnose both cardiovascular and non-cardiovascular diseases to track its progression and to assess the efficacy of therapies. The measurement and interpretation of heart rate and heart rate variability depend critically on how these quantities are computed from the time-series of R-wave events on the electrocardiogram. While the design of algorithms to compute heart rate and to assess heart rate variability is an active area of research, none of the current approaches considers the natural point process structure of human heart beats, together with the physiology underlying the generation of the discrete, biological events.
To address these issues, the first four specific aims of this project are to test the hypotheses that: 1) Human heart beats can be accurately characterized by using a statistical framework based on point process models of the R-R intervals and that this framework can be used to establish new definitions of heart rate and heart rate variability. 2) We can develop local maximum likelihood and point process adaptive filtering algorithms to track in real-time heart rate and heart rate variability and goodness-of- fit methods based on the theory of point processes can be used to assess the agreement between human heart beat series and model estimates of these series derived from the algorithms. 3) The algorithms developed in Specific Aim 2 can be used to construct time domain and frequency domain measures of heart rate variability and to detect and correct, ectopic, erroneous and missed beats in heart beat series. 4) The analysis paradigm developed in Specific Aims 1 to 3 can be used to characterize cardiovascular and autonomic function in tilt-table and autonomic blockade assessments of the cardiovascular system, pathophysiology assessment in congestive heart failure, functional magnetic resonance imaging studies of the brain during meditation, and studies of circadian and sleep physiology. Specific Aim 5 is to provide on our website software to implement the statistical methods developed to address Specific Aims 1 to 4. This will facilitate the research of investigators wishing to characterize heart rate and heart rate variability as part of their physiological studies. The broad, long-term objectives of the project are to provide researchers with a coherent statistical paradigm to characterize cardiovascular control through analysis of heart beat interval dynamics.
The health implications of this project are a more accurate characterization of cardiovascular control in research and clinical studies of both normal and pathological conditions.

The Project

The specific aim of this project is to provide new definitions of heart rate and heart rate variability that could have important implications for research studies of cardiovascular and autonomic regulation and for heart rate monitoring in clinical settings. While the design of algorithms to compute heart rate and assess heart rate variability is an active area of research, none of the current approaches considers the natural point process structure of human heart beats. We model the stochastic structure of heart beat intervals as a history-dependent point process and derive from it an explicit probability density that gives new definitions of heart rate and heart rate variability. Our new methods suggest several new applications and research directions. First, they suggest a new algorithm for simulating heart beats for cardiac pacing. Second, they provide a means to characterize normal and pathologic conditions in terms of heart rate variability, which can be computed along with heart rate in real-time. These analyses may be used to both diagnose disease and follow its progression. Third, our algorithms may be used for real-time monitoring of fetal heart rate variability during antenatal or intra-labor studies, as well as for monitoring patients in the intensive care unit and in the operating room. Fourth, incorporating the point process framework into models of cardiovascular system control and autonomic regulation may enhance their accuracy. In particular, studying cardiovascular control with extensions of these methods to state-space representations of the autonomic nervous system and using point process filtering algorithms for dynamic estimation may help clarify the relation between stochastic and deterministic models of human heart beat dynamics.

  In the first point process model implemented, we derive an explicit probability model for heart rate under the assumption that the stochastic properties of the R-R intervals are governed by an inverse Gaussian renewal model. We estimate the time-varying inverse Gaussian parameters by local maximum likelihood, and assess model goodness-of-fit by Kolmogorov-Smirnov tests based on the time-rescaling theorem. We illustrate our new definitions in an analysis of human heart beat intervals from ten healthy subjects undergoing a tilt table experiment. We report instantaneous heart rate variance signal estimates and show that they provide different information from that in the instantaneous heart rate signal estimates. Our framework gives a more physiologically sound representation of the stochastic structure in heart rate than those provided by current definitions and analysis methods. This work has been published in the American Journal of Physiology: Heart and Circulatory Physiology (Barbieri et al., 2005).

  Based on the model above, we also implemented an adaptive point process procedure to estimate instantaneous time-variant heart rate variability indices, and we demonstrated the ability of our method to track instantaneous dynamics in autonomic regulation of the cardiovascular system in the same tilt table protocol. The adaptive algorithm can update the heart rate variability estimates at any time resolution, obviating the need for interpolation, and can track fast dynamics by considering only the actual information at each time step. The algorithm is easy to implement for on-line analysis of heart rate variability in the intensive care unit, operating room or labor and delivery suits. The dynamics of our indices of heart rate variability may be useful in characterizing normal and pathological conditions of cardiovascular control and regulation. The adaptive algorithm has been published in IEEE Transaction on Biomedical Engineering (Barbieri et al., 2006).

   Our paradigm offers a new set of tools to study autonomic regulation of the cardiovascular system in both research and clinical settings. Current work is focusing on more complex history dependence models for human heart beat generation, and on more complex representations incorporating the point process framework into models of cardiovascular system control and autonomic regulation. In particular, this can be achieved studying cardiovascular control with extensions of these methods to state-space representations of the autonomic nervous system, as well as inclusion of other cardiovascular variables such as arterial blood pressure, central venous pressure, and respiration. More detailed informations about our algorithm and demo programs can be found here.