1Systems and Applications R&D Center, Texas Instruments, Dallas, TX, USA
Ultrasound systems are signal processing intensive. The signal processing for a typical ultrasound system can be roughly divided into three components: the front end which includes digitization of ultrasound signal followed by beam-forming to focus on a particular scan line, the mid end which performs demodulation, envelope detection, compression, and velocity estimation, and finally the back end which performs various image enhancement, rendering functions and user interfaces. In this paper, we describe how the ultrasound signal chain can be efficiently implemented in the class of embedded processors known as Digital Signal Processors (DSP).
We have characterized the performance of various Ultrasound algorithms in several classes of DSP architectures from Texas Instruments (TI). Specifically we have optimized implementations of mid-end and back end processing needed for B-mode and color flow imaging mode. We have used TI’s software framework to put together a demonstration processing systems that shows the capabilities of these devices.
We have demonstrated the operations of ultrasound back-end processing which includes magnitude and velocity estimate from the demodulated data, compression for B-mode magnitude, scan conversion to match the display as well as color flow versus B-mode selection on one of TI’s evaluation modules. With only a 430 MHz DSP, we can perform these operations for 30 frames per second, 96 scanlines per frame, 256 samples per scanline and 10 ensembles for velocity estimate with about 84% loading of the DSP. New multi-core platforms run up to 1.2 GHz with 8 of these cores available for processing on a single device. We are now measuring performance including the mid-end processing across these devices and will report on these performance measures at the conference. A very detailed analysis of the capabilities implement the front end functionalities has been submitted for regular paper in this conference as well. We will show how these devices can be put together to develop the digital part of the ultrasonic system. Some of these devices also include reduced instruction set Computing (RISC) processor as well as display units to handle the necessary control, user interface, display, and saving of data to external devices like hard disks or memory cards.
DSP based low power embedded processors have revolutionalized the compute platforms for consumer electronics. The same technology is well suited for ultrasound systems. The family of devices available to ultrasound system designers using the architectures analyzed in this paper varies from single core to multi-core devices. These devices thus bring the added advantage of scalability. Portable to high end systems can all be based on same baseline embedded architecture and code.