High-Frame-Rate Ultrasound Imaging Innovations for Complex Flow Quantification

Vascular pathology such as atherosclerosis or arterial stiffening involves significant disruption in the blood flow dynamics found in the vasculature, with complex behavior such as jets and swirling typically seen in diseased vessels. Detecting and mapping complex blood flow dynamics in the human vasculature is therefore important for monitoring disease progression and understanding the pathology. Existing solutions for flow imaging in the clinic cannot reliably capture complex blood flow dynamics. These solutions also suffer from limited frame rates, high operational costs and inaccessibility for point-of-care and routine use. To meet this need, this dissertation has devised a novel ultrasound-based imaging framework suited for visualizing and quantifying complex blood flow in human vasculature at the bedside. This was achieved using next-generation ultrasound technology and a novel vector flow imaging framework. The framework utilized deep learning and parallel computing principles to achieve real-time vector flow imaging with an expanded range of measurable velocities. The proposed framework was realized on a programmable and mobile ultrasound scanner to support point-of-care use. Performance of the proposed framework was validated in phantom and human experiments, including healthy and diseased arteries. Reliable vector flow imaging was achieved at the bedside in human trials where complex and high flow speed blood flow was imaged at the bedside. This dissertation presents important progress towards the visualization of complex blood flow dynamics in human vasculature at the bedside and the translation of advanced ultrasound imaging technologies to the clinic. By providing a reliable implementation of vector flow imaging at the bedside, the presented framework should facilitate more comprehensive and large-scale human studies into the hemodynamics of healthy and pathological vasculature. This work also showcases the novel applications of deep learning and parallel computing principles to ultrasound flow imaging, thereby contributing new strategies to the field