Accelerating USG image reconstruction with SAR implementation on CUDA

Abstract

Ultrasonography scanning allows to obtain a cross-section picture of a scanned object, and medical ultrasonography has become one of the most common and safe diagnostic methods. In the process of ultrasonographic image rendition probes with ultrasonic sensors transmit sound waves into an object being scanned and receive the reflected echoes. The received data is reconstructed into a digital image. Since reconstruction algorithms process massive amounts of data, acceptable performance has traditionally been obtained due to dedicated architectures based on specialized integrated circuits and embedded software. In this paper we present a new, fast implementation of USG image rendition based on SAR algorithm efficiently implemented on low-cost and commonly available CUDA architecture. We describe the SAR algorithm, explain its CUDA implementation and demonstrate test results featuring a significant speed-up in rendition time. © 2012 Springer-Verlag.