Ultrasound Measurement System for Three Dimensional Dynamic Structures

提出了一种基于编码传输的动态超声三维成像技术。采用传统的掷骰子填充工艺制备了pzt40x40元件阵列。阵列尺寸为40 mm × 40 mm。由于每个单元的尺寸比较大（1mm × 1mm），为了扩大超声波束，选择中心频率为2mhz。平均分数带宽为59.1%，灵敏度方差约为10 dB。我们确认了为每次传输形成一个3D图像序列的可行性。我们还验证了使用呼吸气球目标的运动估计和补偿。气球的直径约为10 ~ 20毫米。呼吸周期为每分钟8转。计算速度是实际应用中最重要的问题之一。我们构建了一个用于高速计算的pc集群系统。将图像重建软件安装到PC-cluster系统中。高速计算的另一种解决方案是用于图像重建的专用硬件。我们利用fpga设计了一个高速图像重建硬件。硬件的图像重建算法由匹配滤波器组成，该滤波器对从视图方向返回的参考回波波形进行匹配。参考回波波形是预先计算的，并作为滤波器的系数存储在存储器中。匹配的滤波器操作在每个传输周期重复。对于单次发送和接收，由于乘法和加法的次数较少，第一种算法的计算时间比第二种算法短。为了评估计算次数和图像质量，在个人计算机上用C语言实现了该算法。最后讨论了该算法在FPGA上的实现。

Dynamic ultrasound three-dimensional (3D) imaging using coded transmission is developed.An array of PZT 40x40 elements was fabricated with a conventional dice-and-fill process. The array size was 40 mm x 40 mm. Since the size of each element is relatively large (1 mm x 1 mm), the center frequency was selected to be 2 MHz in order to expand the ultrasonic beam. The averaged fractional bandwidth was 59.1%, and the sensitivity variance was approximately 10 dB.We confirmed the feasibility of forming a sequence of 3D images for every transmission. We also confirmed the motion estimation and compensation using a breathing balloon target. The size of the balloon is about 10 〜 20 mm across. The cycle of breathing is 8 rpm.The speed of computation is one of the most important subjects for practical applications. We constructed a PC-cluster system for high-speed computation. The image reconstruction software was installed into the PC-cluster system.An alternative solution to the high-speed computation is a dedicated hardware for image reconstruction. We have designed a high-speed image reconstruction hardware using FPGAs.The image reconstruction algorithm for the hardware consists of matched filters matched for the reference echo waveforms return from view directions. The echo waveforms for the reference are calculated in advance, and are stored in the memory as the coefficients of the filters. The matched filter operation is repeated for each transmitting cycle. For single transmission and reception, the calculation time of the first algorithm is short compared with the second algorithm, because of the less numbers of multiplies and additions.The algorithm was implemented in a personal computer using the C language in order to evaluate calculation times and the image qualities. Implementation of the algorithm by FPGA are also examined.