Sistema de processamento digital de sinais ultrassônicos baseado em system on a chip (SoC)

Abstract

Ultrasound (US) imaging is a widely spread technique, spanning applications from diagnostic medicine to the general industry. Its non-invasive and non-ionizing characteristics make it suitable for various medical exams. Research in this field often explores novel radiofrequency US techniques, which require full access to intermediate signals and the capacity to make changes in processes that are commonly not accessible in commercial US platforms. Open platforms emerge as a solution, facilitating research in the US domain. The proliferation of the System on a Chip (SoC) technology has accelerated advancements in this field, enabling the creation of open systems with malleable characteristics in transmission beamforming, access to Analog to Digital Converter (ADC) raw data, and real-time imaging, either in hardware, software or in hybrid platforms SoCs. This work aimed to implement an open system to process back-end US data in a SoC platform from AMD Xilinx, specifically the ZedBoard, utilizing the reception beamformer and B-mode image reconstruction with scan conversion. Each processing step was modeled with Model Composer blocks in Simulink, implemented utilizing Vivado and Vitis software, and validated with empirical US data acquired from a commercial phantom. The results of intermediate steps were compared with MATLAB/Simulink simulations, utilizing Normalized Root Mean Square Error (NRMSE) and Normalized Residual Sum of Squares (NRSS) metrics to validate the agreement between models. Image results of the system‘s empirical implementation indicate that it serves as a suitable US processing platform with quantitative errors of 7.14% for Contrast-to-Noise Ratio (CNR) and 17.52% for Contrast Resolution (CR), and with a processing time of 0.36 seconds to display the resulting image on a VGA screen, utilizing less than 50% of the available SoC resources. Although the results show a minor degradation in the contrast resolution compared to the reference, they suggest that the resulting images retain essential information. Moreover, the device’s resource utilization indicates potential optimization for additional processing steps, including changes to enhance processing time, which is already faster than previous works. Possible applications of this work are in the further development of future US research open systems to enable innovative research in the ultrasound area.