FPGA-based embedded signal processing for 3D ultrasound computer tomography

We are developing a three-dimensional ultrasound computer tomography (3D USCT) system for early breast cancer diagnosis. The interaction of unfocused ultrasonic waves with the female breast is recorded as pressure variation over time from many different angles and afterwards computationally focused in 3D. For data acquisition (DAQ) a massively parallel, FPGA-based system has been developed. The measurement data is subsequently transferred to a PC for image reconstruction, which is very time consuming. A promising approach in reducing reconstruction time is mapping calculations to the FPGAs in the DAQ system to benefit from parallel processing in hardware. In this work, a set of signal processing steps, deployed for improvement of imaging contrast, was ported to the FPGAs. The suitability of FPGA-based processing was investigated for different chains of computation, i.e. time and frequency-domain processing, with focus on processing performance as well as resource utilization. In the best suited approach, processing time per data set and FPGA was slightly above 50μs, occupying less than 30% of device resources. Thus, the studied processing steps can be done in a pipelined fashion during data transfer, which only gives additional latency and hides processing time.