Modeling and Identification of Peripheral Artery Behavior and Systemic Hemodynamic using Non-invasive Wearable Sensors

Abstract

The state of the peripheral arteries is known to be a key physiological indicator of the body’s response to both acute and chronic medical conditions. For example, the body’s vascular tone, or constriction of the arteries relative to a maximally dilated state, is a direct indicator of the body’s response to cardiovascular stress. However, vascular tone is very difficult to assess using existing technologies, especially noninvasively. Peripheral arterial constriction or dilation is also the dominant factor in determining the body’s systemic vascular resistance (SVR), or resistance felt by the heart in forcing blood through the circulatory system, and vascular resistance change is a major means of compensation to maintain physiological homeostasis. Situations where rapid changes in vascular tone and SVR are known to have great importance include shock (septic, cardiac, traumatic, etc.), post-surgical recovery, and hemodialysis. The candidate proposed a technique for tracking changes in vascular tone by combining a photo plethysmography sensor with an adjacent compliant piezoelectric polymer pressure sensor (polyvinylidene fluoride). A simple local model for viscoelastic dynamic behavior of the underlying artery and surrounding tissue is generated and coupled to the piezoelectric sensor model, from which variations in relative amplitude and hysteresis between the piezoelectric and photo plethysmograph signals are found to show strong correlations with invasively measured SVR data in swine subjects. The mean absolute percentage errors were less than 4.7% and root mean square errors were less than 0.037 for all three swine subjects. A local nonlinear artery model with extended Kalman filter performed system identification and tracking of the radius of the peripheral arteries as well as blood pressure. In proof-of-concept testing on a swine test subject, local vascular resistance calculated from arterial radius estimates at the ring location showed good agreement with overall systemic vascular resistance, with a 2.7% mean absolute percentage error and 0.026 root mean square error, while capturing other features of local cardiovascular behavior more noisily. Further validation is performed with ultrasound measurements of foreleg arterial radius while measurements with compliant sensors are taken. Additionally, the candidate introduced a new systemic hemodynamic model, combining vascular resistance with heart rate, which may provide substantial insight into cardiovascular response to clinical interventions. This work attempted to better understand how estimated changes in local peripheral arterial radius obtained from wearable sensors relate to dynamic compensation in the full cardiovascular system. Preliminary human study results show that hemodynamic decompensation can be predicted under criteria based on peripheral vascular resistance from local EKF estimations and systemic hemodynamic feedback model error. Under the certain criteria, sensitivity to future decompensation was 100% and specificity was 88% for the training data set, and 100% and 84% for the full 50 patient sample.