3-D Super-resolution ultrasound imaging using deep learning approach
Super-resolution ultrasound imaging is an emerging technology capable of visualizing microvasculature at micron-level spatial resolution. While successfully applied in various preclinical studies to assess abnormal vascular alterations associated with disease progression, most current techniques produce only 2D cross-sectional images, unable to reconstruct complete 3D vascular networks. Our team has successfully demonstrated the feasibility of 3D volumetric super-resolution imaging using a 2D ultrasound array. However, there is a room to further advance the technology, particularly in addressing large data rates, low frame rate, and long scan times. This project aims to utilize deep learning approaches to further advance 3D super-resolution technology. The motivated student will have the opportunity to enhance various aspects of the technology, including ultrasound beamforming, signal processing, and GPU implementations. The improved technology will be applied to different animal models, potentially making a broad impact across various research projects.
Investigation on an innovative ultrasound image reconstruction: quality enhancement, sub-sampling, and deep learning
Ultrasound imaging used relies on an ultrasound probe that consists of an array of many transducer elements and electrical channels. The array transducer requires beamforming technologies to acquire and form the images. Various technologies have been developed for better image acquisition methods and reconstruction algorithms. Acquisition and reconstruction of high-quality images for practical uses require a high resolution, a large field of view, a high frame rate, a less data sampling, a reduced hardware (less electronic signal channels), especially for the technology advancement in 3-D volumetric imaging. However, given the state-of-the art hardware, these requirements are usually limited by currently known physics. In this project, a deep learning approach is proposed to explore an unconventional ultrasound image reconstruction. The motivated students can work independently or/and collaboratively to learn and explore different technical approaches to achieve this goal. They can learn and perform algorithm development, computer simulations, and ultrasound imaging experiments.
Human cardiac biomechanics and novel ultrasound structural ventricular imaging
Pulmonary hypertension (PH), or chronic high blood pressure, is a widespread health problem with no known cure other than lung transplantation. Mid- to late-stages of the disease are characterized by mechanical and structural alteration to the muscles that compose the walls of the heart chambers. However, the precise nature of these changes in humans is not known. Further, current clinical strategies for risk assessment in PH do not include structural or functional imaging metrics, instead relying on invasive catheterization methods. Our group is investigating how human heart muscle changes mechanically and structurally during PH, and simultaneously developing novel ultrasound imaging technology to monitor these changes in a clinical setting. The interested student will elucidate these outstanding and impactful questions by characterizing mechanical and structural remodeling of human heart muscle using benchtop biomechanical testing and finite element modeling of excised tissues, and by further developing the novel ultrasound imaging technology.