Use this URL to cite or link to this record in EThOS:
Title: Super-resolution ultrasound imaging with microbubbles
Author: Christensen-Jeffries, Kirsten Mia
ISNI:       0000 0004 6497 8497
Awarding Body: King's College London
Current Institution: King's College London (University of London)
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Ultrasound imaging is one of the most widely used clinical imaging methods offering safe, real-time imaging at low cost with excellent accessibility. However, the structure and flow of deep microvasculature, which can serve as a marker of pathological or dysfunctional tissue, cannot be adequately resolved using standard clinical ultrasound imaging frequencies due to diffraction. Conventional ultrasound imaging resolution is related to the wavelength employed, however, high frequency approaches used to improve resolutions are limited in penetration depth. Therefore, there is a crucial clinical need for the development of new techniques that can fill this ‘resolution gap’. This work develops a technique to generate super-resolved images of the vasculature using accumulated localisations of spatially isolated microbubble contrast signals. Furthermore, a temporal tracking algorithm is introduced, enabling the extraction of fluid flow velocities. Using this approach, in vitro flow phantoms are visualised to a depth of 7 cm at sub-diffraction scale using standard clinical ultrasound equipment. In subsequent work, super-resolution imaging and velocity mapping are demonstrated in vivo, providing quantitative estimates of blood flow velocities at a super-resolved spatial scale. The algorithm is then extended to acquire quantitative measures for the clinical evaluation of human lower limb perfusion, where super-resolution localisation measures are able to identify differences in the microcirculation between patients and healthy volunteers following exercise. Super-resolution imaging relies on the correct identification of spatially isolated bubble signals, while user defined thresholding limits its clinical translation. To address this challenge, machine learning techniques for foreground detection and signal classification are investigated. It is shown that support vector machines provide promising results for super-resolved imaging, whereas the unsupervised approaches investigated appear unsuitable. In addition, the 2D acquisition strategy employed limits the application of the technique to structures with limited 3D complexity. This work concludes by developing a fast, multi-probe approach, which allows 3D super-resolution imaging and flow detection in vitro.
Supervisor: Eckersley, Robert Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available