Candidate: Hassan Nahas
Title: High-Frame-Rate Ultrasound Imaging Innovations for Complex Flow Quantification
Date: November 23, 2020
Time: 2:00 PM
Place: REMOTE ATTENDANCE
Supervisor(s): Yu, Alfred
Abstract:
Blood flow dynamics carry an immense amount of diagnostic information and are the basis of many of the biomarkers used in diagnosing and assessing cardiovascular diseases. For example, blood flow reaches high speeds and complexity in the vessels affected by vascular disease, therefore detecting and quantifying high-speed and complex flow is necessary for characterization and diagnosis. Currently, ultrasound is our best tool for noninvasively visualizing blood flow in vivo as alternative modalities such as Magnetic Resonance Imaging suffer from limited temporal resolution and high operational costs. Nonetheless, conventional ultrasound cannot capture all the details of cardiovascular flow, especially the multi-directional high-speed dynamics that occur in pathology. Past efforts that attempted to overcome the limitations in conventional ultrasound imaging using advanced techniques are not robust enough for quantitative analysis. They suffer from significant biases and imprecision, particularly in conditions with complex flow such as vascular disease. This makes any quantitative analysis unreliable and limits the clinical value of these techniques. Furthermore, these advances have come at a significant cost in processing time, preventing their application on-site. On the other hand, the ideal imaging modality is a realtime point-of-care solution. To meet this need, I will devise a novel bed-side imaging framework suited for visualizing and quantifying the high-speed multidirectional flow in vivo, particularly as they emerge in conditions of vascular stenosis for large arteries. I hypothesize that this can be realized using next-generation ultrasound flow imaging techniques augmented with three new innovations: (1) Deep learning to identify errors that corrupt flow estimation using ultrasound; (2) Robust signal processing to correct these errors and derive flow velocity in complex scenarios; (3) Parallel computing to achieve real-time processing for the entire framework on the bedside. Throughout the proposed timeline, validation of the developed vector flow estimation techniques will be performed using in vitro flow models with controllable flow dynamics, as well as an in vivo scenarios. The advancements made in the proposed research towards robust vector flow estimation will ensure the reliability of the quantitative biomarkers generated for assessing cardiovascular disease and predicting clinical outcomes. Through this work, I will also advance the role of machine learning in signal processing for ultrasound imaging, which is currently the subject of great interest in the community. Moreover, the proposed research will contribute significantly to the translation of advanced ultrasound imaging technologies to the clinic, as a bed-side high-speed vector flow estimator has yet to be realized. With such as tool in place, research and clinical care of cardiovascular health should become more comprehensive and efficient.