Challenging the multimodal ultrasound with GPU-based research ultrasound platforms

Old-school ultrasound was all about discrete modalities (B-mode, Doppler). Today, we prefer to combine multiple methods and advanced processing algorithms to extract more diagnostic data out of the raw RF ultrasound echoes. Thus, high-speed acquisition and processing of channel data have become a standard feature of contemporary research systems. I will present our developments in the programmable and scalable ultrasound systems featuring GPU-based processing. The implementation of multimodal methods is enabled through a high-performance data streaming and software processing approach. We will look into GPU processing and how challenging real-time realization of the Software Defined Ultrasound paradigm can be. We will also discuss the many new opportunities and open-source tools the software approach nevertheless opens up, which can be used for both standard ultrasound processing and advanced post-processing (e.g., machine learning).The case-studies presented show how the platforms can be used as versatile research tools, as well as demonstrators for commercial introduction of the technology. A sneak preview of a 3-D-ready research system will conclude my talk.