Real-Time Embedded Implementation of Adaptive Beamforming for Medical Ultrasound Imaging

Medical ultrasound diagnoses have been applied to a variety of clinical occasions for years. Recently, with the increasing needs for mobile and portable medical diagnoses, the development of medical ultrasound imaging algorithms on embedded platforms is flourishing. Among various medical ultrasound imaging algorithms, the typical delay-and-sum beamforming algorithm is a simple algorithm which is easy to realize a real-time embedded imaging implementation. However, this beamforming algorithm has relatively low image quality. As such, the implementation of high-quality minimum-variance adaptive beamforming algorithm for medical ultrasound imaging was studied on heterogeneous embedded computing platform which contained a high-performance embedded GPU. By applying the efficient implementation strategies, the embedded GPU implementation of the adaptive beamforming algorithm performed more than 100 times faster than its ARM processor implementation counterpart. Furthermore, scalability and power consumption of the embedded adaptive beamforming implementation were also evaluated, which demonstrated that the heterogeneous embedded computing platform utilized in this paper was suitable for a real-time mobile or portable high-quality medical ultrasound imaging device construction.