Implementing subarray average delay multiply and sum ultrasound beamformer using OpenCL

Abstract

The subarray average delay multiply and sum (SADMAS) beamformer, proposed by our group recently, has achieved significant improvement in the resolution and performance of the ultrasound images with lower computational complexity compared to MV or DMAS beamformers. Since SADMAS beamformer is based on matrix processing, it has the advantage to be used in ultrasound system based on FPGA or GPU for real-time imaging. In this paper, we put forward an implementation of the SADMAS beamformer on a high performance graphics processing unit (GPU) by using Open Computing Language (OpenCL). In this implementation, the specific algorithm architecture was designed and implemented by 2 kernel-based parallel beamformer. To evaluate the performance of the algorithm architecture, point phantoms were used to evaluate the performance of the algorithm. The execution time and the frame rates were measured by using this implementation targeted on GPU and CPU, respectively. Experiment results show that, by using different subarray length with M/2 and M/4 (where M is the size of array), the frame rates improve 38.6 and 44.6 times, respectively, compared with the beamformer implemented on CPU. Those results indicate that the implementation of SADMAS beamformer on GPU is feasible for real-time imaging in medical ultrasound system.