Chen Zhaohong, Chen Yongdong, Huang Qinghua
IEEE Trans Biomed Circuits Syst. 2016 Apr;10(2):394-403. doi: 10.1109/TBCAS.2015.2420117.
Ultrasound elastography is an important medical imaging tool for characterization of lesions. In this paper, we present a wireless and near real-time 3D ultrasound strain imaging system. It uses a 3D translating device to control a commercial linear ultrasound transducer to collect pre-compression and post-compression radio-frequency (RF) echo signal frames. The RF frames are wirelessly transferred to a high-performance server via a local area network (LAN). A dynamic programming strain estimation algorithm is implemented with the compute unified device architecture (CUDA) on the graphic processing unit (GPU) in the server to calculate the strain image after receiving a pre-compression RF frame and a post-compression RF frame at the same position. Each strain image is inserted into a strain volume which can be rendered in near real-time. We take full advantage of the translating device to precisely control the probe movement and compression. The GPU-based parallel computing techniques are designed to reduce the computation time. Phantom and in vivo experimental results demonstrate that our system can generate strain volumes with good quality and display an incrementally reconstructed volume image in near real-time.
超声弹性成像技术是一种用于病变特征表征的重要医学成像工具。在本文中，我们展示了一种无线且近实时的三维超声应变成像系统。该系统使用一个三维平移装置来控制商用线性超声换能器，以采集预压缩和后压缩射频（RF）回波信号帧。这些RF帧通过局域网（LAN）无线传输到高性能服务器。在服务器的图形处理单元（GPU）上，利用计算统一设备架构（CUDA）实现动态规划应变估计算法，以便在接收到同一位置的预压缩RF帧和后压缩RF帧后计算应变图像。每个应变图像被插入到一个应变容积中，该应变容积可以近实时地渲染。我们充分利用平移装置精确控制探头的移动和压缩。基于GPU的并行计算技术旨在减少计算时间。仿体和体内实验结果表明，我们的系统能够生成高质量的应变容积，并近实时地显示增量重建的容积图像。
