Physiologic Signal Processing and Modeling

Ramakrishna Mukkamala | rama@egr.msu.edu | www.egr.msu.edu/people/profile/rama

 

Research in the Physiologic Signal Processing & Modeling Laboratory includes cardiovascular physiology, mathematical modeling, physiologic measurement, signal processing, system identification.

Hemodynamic Monitoring by Blood Pressure Waveform Analysis

Blood pressure waveform analysis represents a potential, practical approach for achieving sorely needed reliable, automated, and less invasive monitoring of hemodynamics.  As a result, investigation of this approach has been longstanding.  However, the previous techniques have neglected key aspects of the physiology and are therefore only able to monitor a limited number of variables that show accuracy over a narrow hemodynamic range.  We have developed a suite of blood pressure waveform analysis techniques that account for the crucial facets of the physiology omitted hitherto via diverse (black-box to physical) models to estimate various essential hemodynamic variables (e.g., cardiac output, left atrial pressure, ejection fraction) from more readily available blood pressure waveforms.  We have demonstrated the validity of these techniques against manual or more invasive reference measurements from experimental subjects and patients during wide hemodynamic perturbations.  Some of these techniques are currently being commercialized by Retia Medical, LLC, East Lansing, MI. 

Advanced Blood Pressure Measurement

Hypertension detection and control rates are unacceptably low, especially in low resource settings.  Advanced blood pressure measurement technology is essential to alleviate this problem.  We have developed a technique to improve automatic cuff blood pressure measurement.  In contrast to current devices, the technique affords patient-specific blood pressure measurement via physical modeling of oscillometry.  We recently showed that the technique was significantly more accurate in subjects with large artery stiffening (a common condition that occurs with aging and disease) and significantly more repeatable than market leading office devices.  We are also currently pursuing cuff-less blood pressure measurement based on the pulse transit time principle (see Figure).  Our innovative ideas in this popular, but challenging, area include the use of ultra-convenient physiologic sensing and modeling to both accurately estimate pulse transit time and to conveniently calibrate pulse transit time to blood pressure. 

Probing Baroreflex Function by System Identification

The baroreflex was long believed to regulate blood pressure only on the time scales of seconds to minutes.  However, experimental studies have now indicated that this system can contribute to long-term blood pressure maintenance.  So, the baroreflex could actually play a causative or protective role in hypertension and heart failure.  It is therefore important to quantitatively probe the baroreflex.  We have developed a number of practical techniques to quantify the functioning of the baroreflex and other neural cardiovascular regulatory mechanisms during normal, closed-loop operating conditions using parametric system identification.  More recently, we employed the Gaussian white noise approach in conjunction with nonparametric system identification to reveal the structure of baroreflex nonlinearity and to discover that the nonlinear baroreflex gain is enhanced in a popular chronic hypertension model even though the linear baroreflex gain is preserved.