Ultrasound-based pulse wave imaging for early detection of arteriosclerosis

Abstract

This thesis presents a study on the design of an ultrasound-based arterial stiffness estimation framework by measuring the localized pulse wave velocity (PWV). Motivation: Arterial stiffening is directly attributed to top mortality causes worldwide; in clinical practice, arterial stiffness is assessed by measuring global PWV across the arterial tree to provide a general impression on the vascular health condition. Global PWV measurements however provide limited indicative information for the onset of arteriosclerosis and is incapable of detecting focal vascular diseases such as atherosclerosis. This has driven the motivation to develop new localized PWV measurement framework for estimation of region specific arterial stiffness to facilitate early detection of arteriosclerosis performed as a clinical routine. Theoretical contribution: Inaccuracy of local PWV estimation using ultrasonography is ascribed to the limited data samples and influence of reflected waves. To address this, a new pulse wave imaging framework is devised by leveraging on broad-field ultrasound imaging principles to increase acquisition rate to better describe the propagating pulse wave, hence reducing estimation variance. The increased signal-data length enabled the development of a signal-processing level mechanisms to suppress estimation errors arising from reflected waves. The filter is designed to decompose Doppler signal vectors into signal subspace components using Eigen-principles with PWV estimated from Eigen-components corresponding to primary pulse wave, disregarding effects of reflected waves. Experimental validation: To address the lack of mature experimental platform for vascular imaging studies, part of this study is devoted to devise a framework for fabrication of ultrasound-compatible thin-walked phantoms to validate stiffness estimates. The PVA-based vascular phantoms are modeled after anthropomorphic structures with capacity to incorporate diseased conditions such as stiffening and stenotic narrowing by manipulating geometrical structures and mechanical properties. Phantom properties are characterized prior to calibration of arterial stiffness estimation techniques. In addition, phantoms exhibiting wave reflection properties are designed to assist in the development of reflected wave suppression filter.

Computational framework design: Implementation of Eigen-based signal processors in real-time ultrasonography is challenging due to its high computational complexity. As such, a multi-level parallel Eigen-computation framework suited for ultrasound signal processing is formulated for fast execution of Eigen-based techniques on parallel computing platforms. To demonstrate real time processing capabilities. The framework was developed for the implementation of a fast-parallelized Eigen-filter for clutter filtering in color flow imaging (CFI) due to the widespread use of CFI in clinical setting.

To summarize, a framework consisting of innovations in data acquisition scheme, signal processing novelties, experimental validation and computational strategy is developed to measure localized PWV. As an impact, this holistic arterial stiffness estimation framework is expected to pave for early diagnosis of arteriosclerosis as a clinical diagnostic routine.