Human-on-the-loop Control for Small Ultrasound Imaging

Ultrasound is widely accepted as one of the best forms of medical imaging compared to similar technologies, such as CT scans and MRI, due its low operating cost and safety for the patients. At the same time, it is well known that there can be large variability in image quality obtained by different experts imaging the same patient. Therefore, new imaging methods are necessary that will decrease this variability and at the same time allow the users to best utilize their expertise. The goal of this project is to develop a new active ultrasound system where expert users interact with a smart ultrasound device in order to improve medical imaging and facilitate diagnosis. The focus is on ultrasound elastography, which involves the quantification of mechanical properties (e.g. elastic or viscoelastic material parameters) in soft tissue. Elastography has gained high prominence in recent years as an imaging technique for non-invasive tissue characterization as mechanical properties have proven to be strong differentiators of disease. Essentially, this method provides a way for non-invasive palpation, a technique that is commonly used by doctors to determine tissue pathology, e.g., for breast cancers. At the same time, ultrasound imaging can be subject to large noise levels, image artifacts, and large image variability when executed by different experts. The latter limitation is due to the large dimension of the ultrasound parameter space that needs to be controlled by the user to obtain good images. While these limitations give rise to the need for automation, the algorithmic complexity of generating elastography maps still necessitates the experience and skills of human operators that allows them to accurately evaluate imaging results and provide guidance for control. The goal is the development of an active ultrasound system where user expertise is used to refine the control process, while autonomous elasticity (or viscoelasticity) mapping improves image quality and allows human users to best use their skills for both optimization and diagnosis. To realize the proposed research, the following four research thrusts need to be addressed: (i) data fusion for ultrasound elastography; (ii) methods for interactive ultrasound elastography; (iii) framework for safe and efficient device implementation; and (iv) validation using laboratory phantoms. By introducing automation in the image collection process and by combining it with user expertise we can obtain ultrasound devices that have significant performance gains compared to present systems that are only manually controlled. Wide availability of such systems can have a significant societal impact in accurate, safe, and cost-effective diagnosis of many medical conditions, such as cancers or liver fibrosis.