BME PhD Defense: Ge Zhu: "Physics-Informed Arterial Tonometry for Noninvasive Continuous Blood Pressure Monitoring and Arterial Stiffness Estimation”

PhD Dissertation Defense  

"Physics-Informed Arterial Tonometry for Noninvasive Continuous Blood Pressure Monitoring and Arterial Stiffness Estimation” 

Ge Zhu 

Abstract: The increasing demand for accurate, continuous, and noninvasive arterial blood pressure (BP) monitoring has motivated the research presented in this dissertation. Continuous BP waveform acquisition provides critical insight into cardiovascular dynamics, autonomic regulation, arterial biomechanics, and hemodynamic responses under both resting and dynamic physiological conditions. Despite significant advances in wearable sensing technologies, existing noninvasive BP monitoring approaches remain limited by frequent calibration requirements, susceptibility to motion artifacts, reduced accuracy during rapid hemodynamic changes, and limited physiological interpretability. 

To address these challenges, this dissertation investigates a physics-based framework for continuous noninvasive BP monitoring centered on arterial tonometry. The superficial temporal artery (STA) was selected as the primary measurement site due to its favorable anatomical characteristics, including shallow depth, underlying bony support, and accessibility for long-term sensing. Building upon these advantages, a novel sensing modality termed superficial temporal arterial tonometry (STAT) was developed. STAT incorporates a biomechanics-informed transfer function that relates externally measured arterial deformation to intraluminal blood pressure. A custom-engineered tonometry device, integrated with dedicated signal acquisition and processing hardware, was designed to enable continuous BP waveform monitoring while reducing calibration dependency and improving measurement stability. 

To further improve physiological fidelity and enable patient-specific modeling, a multimodal sensing framework combining ultrasound shear wave elastography (SWE) and arterial tonometry was developed. This integrated system enables simultaneous acquisition of arterial stiffness and pulse waveforms, providing subject-specific biomechanical parameters for enhanced BP estimation. To support controlled evaluation of the proposed methods, a programmable closed-loop pulsatile flow phantom platform was also developed. The platform reproduces physiologically relevant hemodynamic conditions through tunable arterial stiffness, pulse pressure, heart rate, and waveform morphology, while maintaining compatibility with clinical-grade arterial line measurements for validation against ground-truth intraluminal pressure. 

In addition, this dissertation introduces a physics-informed neural network (PINN) framework for estimating patient-specific arterial stiffness directly from tonometry-derived pulse waveforms. By embedding biomechanical constraints within the learning process, the proposed PINN improves both interpretability and generalizability relative to purely data-driven approaches. This capability extends tonometry beyond BP estimation alone, enabling extraction of physiologically meaningful cardiovascular biomarkers from wearable-compatible measurements. 

The performance of the STAT system was evaluated in a human pilot study involving ten participants across twenty-nine experimental sessions, each consisting of thirty minutes of continuous monitoring during both resting conditions and isometric handgrip exercise. STAT-derived BP waveforms were collected concurrently with measurements from the Finapres Finometer, serving as the reference standard for continuous noninvasive BP monitoring, and were additionally compared against a widely used cuffless BP estimation approach based on pulse transit time (PTT). The proposed STAT system achieved a mean absolute difference of 4.8 ± 2.2 mmHg under resting conditions and 6.5 ± 3.4 mmHg during dynamic conditions, demonstrating improved capability for tracking rapid BP fluctuations compared with optical based-based cuffless BP monitoring methods. 

Collectively, this dissertation establishes a comprehensive framework for continuous noninvasive cardiovascular monitoring through the integration of biomechanics-based arterial tonometry, multimodal elastography sensing, physiologically realistic benchtop validation, and physics-informed machine learning. The contributions presented herein advance the development of personalized, calibration-light, and wearable BP monitoring technologies while providing new opportunities for physiologically interpretable cardiovascular assessment and precision health applications. 

For a zoom link, please email kharrison@wpi.edu