UC San Diego

Ultrasonography has been widely applied in imaging tissues and organs for diagnosing diseases due to its capability of serving as a real-time, non-invasive, portable, and radiation-free tool. Ultrasound imaging has demonstrated significant value in diagnosing and monitoring periodontitis that can improve the limitations of conventional methods such as radiography and periodontal probing. While radiography provides excellent sensitivity to hard tissues, it uses ionizing radiation and is limited in stratifying disease in soft tissues. Periodontal probing is time-consuming and painful for patients, and inter-operator variation in probing can exceed 40%.In this dissertation, the first part reviews the materials and mechanisms of ultrasound transducers. The second part includes our work that developed a wearable skin-conformal ultrasonic phased array patch for continuous haemodynamics monitoring in deep tissue. The penetration depth is up to 14 cm beneath the skin, allowing accessing biological signals from central organs. The third part includes our work that developed a toothbrush-shaped ultrasound transducer to image all 28 human teeth with its miniaturized size and side-facing design. The results showed that the full-mouth imaging process by the miniaturized transducer is reproducible in a small cohort of five subjects. Besides imaging anatomical structures, blood flow of the supraperiosteal artery in human gingiva was measured by color and spectral Doppler to provide hemodynamic information. The fourth part includes our work that mapped the location and physiology of the greater palatine artery (GPA) in a 3D manner. Mapping the trajectory of the GPA in the hard/soft palate is clinically important to evaluate the available palate donor tissue during soft tissue autografting and to avoid sectioning of the GPA during the surgery. The fifth part summarizes our work of reconstructing sparse array ultrasound imaging with deep learning algorithms. Sparse array ultrasound imaging has presented potential in reducing the fabrication cost, energy consumption and electrical system complexity. However, sparse array imaging suffers from inevitable artifacts due to grating lobes. We developed a convolutional neural network (CNN)-assisted sparse array using only 16 channels out of a 128-channel transducer, in which the pitch is eight times that of the dense array.