Timing estimations of cardiovascular events; applications to seismocardiography, microneurography, and blood pressure.

There is a wealth of untapped information within commonly acquired cardiovascular signals. Electric, vibration, and pressure measurements in this system allow us to obtain precise timings of events that can inform us about its ability to maintain health. In particular, this thesis applies new data analysis methods to continuous blood pressure, muscle sympathetic nerve activity, and seismocardiography. Three phase estimation techniques, including one of our own design, were compared in terms of accuracy. The three techniques were based on wavelet analysis, Hilbert analysis, and a new peak detection method, respectively. They were applied to modelled in-silico data, as well as a set of four pairs of in-vivo data. The results showed that the wavelet method should be selected for data with signal-to-noise ratio above −5 of unknown or varying frequency. Otherwise, all three techniques performed with equivalent accuracy, with the wavelet technique being computationally slower. The new peak detection technique was applied to blood pressure and muscle sympathetic nerve activity data on participants undergoing lower body negative pressure. In this sit- uation, the peak detection method offered better time resolution, and did not make the assumption that the signals were composed of a sum of sine waves. New indices were developed to identify timings within each time series, and quantify the relationship between the two signals. The indices returned values analogous to those obtained from traditional methods. One index differentiated between the participants that completed the lower body negative pressure protocol without exhibiting symptoms of pre-syncope, and those participants who did not complete the protocol. A third study considered seismocardiography, which measures thoracic vibration caused by the beating heart, and contains unique information about cardiac mechanics. Starting from a wavelet analysis basis, an algorithm capable of obtaining precise timings in seismocardiogram signals without the use of any other concurrent signal was developed. The algorithm included a new model of seismocardiogram systolic vibrations, fitted by minimizing an original distance function. At levels of lower body negative pressure of intensity below −30 mmHg, the algorithm was 95% accurate, and the heart rate variability indices were not statistically different from those obtained with electrocardiography.Cardiac timings that are represented in seismocardiogram peaks include isovolumic mo- ment, aortic valve opening, and aortic valve closure. It is known that the valve opening and closing peaks are not caused by the movement of the valve itself. Furthermore, the isovolumic moment peak, defined as the seismocardiogram minimum that occurs during the isovolumic contraction, does not correspond to any known, precisely timed event. In fact, the mechanical processes that cause seismocardiogram fiducial points have not been identified. Two 3D meshes were developed to study the propagation of vibrations in the thoracic cage. The first mesh was created from basic geometrical shapes and the second was adapted from a life-like full-body human mesh. By modelling the viscoelastic proper- ties of materials therein, we applied a previously developed solving algorithm to simulate seismocardiograms caused by a heart-like force applied to the sternum. Both simulations contained peaks analogous to all in-vivo seismocardiogram fiducial points.