Seminars

External Reviewer: Dmitry Grishenkov ( Associate Professor at Department of Biomedical Engineering and Health Systems, KTH - Royal Institute of Technology)

Abstract:
Ultrasound (US) is a widely used medical imaging modality due to its low cost, non-ionizing nature, and real-time capabilities. Typical B-mode US images display the reflectivity of tissue structures, which however is only qualitative information. Speed-of-Sound (SoS) is a novel quantitative biomechanical marker, which can provide complementary and/or independent information on tissue composition and pathology, e.g., for diagnosing breast cancer and liver steatosis. In my thesis, I study novel methods for SoS estimation using conventional hand-held transducers.

In this talk, I will first present a method for estimating a single representative SoS value for the entire imaged tissue. This allows for biomechanical characterization, e.g., for breast density classification that is valuable as breast cancer risk factor. Next, I will talk about image reconstruction methods for mapping local SoS distribution, based on observed echo shifts from different viewing directions. We solve an inverse problem of image reconstruction based on a forward imaging model. Such forward models are typically hand-crafted based on the understanding of involved medical physics, electrics, and computational phenomena. We introduce a novel approach for learning such reconstruction forward model based on data. I will also present SoS imaging using model-based deep learning via loop unrolling as well as its pruning through distilled priors.