Research On Ultrasonic Elastic Imaging Method Based On GPU Method

Biological tissue’s movement/displacement estimation and strain estimation are hotspot in ultrasound elastography. Recently, in order to estimate displacement, researches have proposed normalized cross-correlation and phase zero search methods; and a low pass differential filter, SG-I have been used to estimate strain. However, the computing performance of these two imaging methods is not high enough, it can’t meet the requirements of real-time and clinical applications. In this paper, we reference parallel thinking to accelerate based on normalized cross-correlation and phase zero search elasticity imaging technology, propose using Jacket on the GPU to parallelize algorithm. First, analyze the details, characteristics and processes of these two elastography technology, we can obtain the parallelization possibilities of the two algorithm’s compute-intensive part. Then, through the GPU development tools Jacket to parallelize the compute-intensive part of the two algorithms. Through simulation human tissue, before and after compression RF signal data, to experiment tested GPU-based both elastography method. The results show is GPU-based method can greatly improve the efficiency of the algorithm. In the case of single-frame signal, based on the cross-correlation method, it can achieve highest speedup366x, and based on phase zero search elastography, the speedup can reach5x.