High-frame-rate volumetric ultrasound imaging using dedicated arrays and deep learning

Abstract

High-frame-rate volumetric ultrasound imaging is highly desired to enable novel clinical ultrasound applications. However, realizing high-quality volumetric ultrasound imaging at a high frame rates (>500 Hz) is challenging. Keeping the cable count and data rate of the transducer device at a realistic level without sacrificing image quality to an undesirable extend means that a dedicated design with carefully chosen trade-offs is required and powerful processing of the received signals is desired. This thesis describes the development of a high-frame-rate 3D ultrasound transducer through dedicated transducer design and explores the use of deep learning-based beamforming to achieve high-quality 3D imaging. Specifically, the first part of this thesis focuses on the development of an imaging scheme and the realization and testing of two prototype transducers for high-frame-rate 3D intracardiac echography (3D-ICE). The second part of the thesis implements deep learning in the image reconstruction process to improve the image quality of volumetric ultrasound. Deep learning-based beamforming is implemented and evaluated first for a miniature matrix array, which similar to the 3D-ICE design applies micro-beamforming to achieve cable count reduction and finally for a spiral array which uses a sparse distribution of transducer channels.