Ultrasound Methods for Quantitative Edema Monitoring

摘要

Patients with end stage renal disease typically must undergo regular dialysis treatments to replace the loss of kidney function. A critical part of these dialysis treatments is the careful management of fluid status, as these patients are at an increased risk for developing fluid overload, a condition that poses a number of dangers to their health and quality of life. Current clinical methods are lacking in their ability to accurately provide a quantitative metric for grading edema and fluid overload. In this dissertation, I explore a number of methods based on ultrasound strain imaging, ultrasound viscoelastography, and ultrasound poroelastography to address this clinical need. The practical and theoretical aspects of the measurement process and parameter estimation methods are explored, and new methods are proposed and evaluated to overcome common difficulties. Chiefly, the experiments and simulations described in this work aim to highlight the role of assumptions in visco- and poroelastic imaging, to explore how these assumptions can hinder accurate parameter estimation, and to develop methods that are less assumption-dependent. First, I evaluate a point-of-care ultrasound viscoelastography system and use it to estimate the viscoelastic properties of a tissue-mimicking material. The strain and material properties are observed to be depth dependent, highlighting possible breaks with the viscoelastic model assumptions and possible poroelastic behavior. Next, I analyze the role of model assumptions on poroelastography measurements using both finite element models and benchtop experiments. Strain magnitudes and loading geometries that differ from the model assumptions used in most poroelastography studies are shown to produce large differences in poroelastic parameter estimates. Furthermore, they can lead to lateral-to-axial strain ratio measurements that do not converge to the true Poisson's ratio of the material, thus highlighting the need for more careful interpretation of standard effective Poisson's ratio (EPR) poroelastograms. Finally, I develop and evaluate a new approach to poroelastography by posing the poroelastic imaging as an inverse problem. This allows for the quantitative imaging of spatial variations. This method is shown to produce more accurate poroelastic images in simulations with ideal, Gaussian corrupted data. In addition, the method shows promise in reconstructions based on simulated ultrasound images, though some difficulties remain. Possible improvements and recommendations for future poroelastography studies are discussed.