带有预测的二维超声射频信号快速位移估计方法的研究

Research on the Fast Displacement Estimation of Two Dimensional Ultrasound RF Signals with Prediction

在超声心肌弹性成像技术中,对采集的超声射频(RF)信号数据进行偏移量化分析是获得准确弹性数据的前提。偏移量化分析中常用的互相关二维位移估计存在计算速度慢、对去相关噪声敏感的缺陷。组织运动是连续的,因此相邻采样点的位移是相似的。利用这一特点,本文提出了一种预测互相关算法,即在进行位移估计时,先对位移进行预测作为初始位移,然后在初始位移基础上,进行小范围搜索,而不必每次都在大范围内进行搜索。实验结果证明,预测互相关算法比传统互相关算法速度提高近3倍,并改进了位移估计精确性。种子RF线位移校正算法的实验结果表明,位移生长法效果要优于梯度均值替换法,这是由于梯度均值替换法无法保证所用的位移是准确的,而位移生长法能确保所用的替换位移是准确的。种子RF线两种位移预测方法比较的实验结果证明,邻域相关系数判断均值法性能要优于直接邻域均值法,这是由于邻域相关系数判断均值法在利用位移进行预测时,对其准确性进行了判断,避免了将错误位移传播下去,提高了位移估计的精确性。

In myocardial elastography, the quantitative displacement analysis of Frequency Radio（RF） signals is the precondition of obtaining accurate elasticity data. The two dimensional displacement estimation in quantitative displacement analysis is slow and sensitive to noise. Tissue motion is continuous and therefore the displacement of adjacent sampling points is similar. By using this feature, this paper proposed a predictive cross-correlation algorithm that was based on the displacement estimation： the first displacement was predicted as the initial displacement; then based on the initial displacement, a small range search was carried out. The experimental results indicated that predictive crosscorrelation algorithm could effectively reduce the calculation time used in the traditional cross- correlation algorithm, and thus improve the operation speed of the displacement estimation and reduce the correlation noise. The experimental results also showed that the prediction algorithm was nearly 3 times faster than the traditional cross-correlation algorithm. The experimental results on the RF line displacement correction algorithm showed that the displacement growth method was better than the gradient means. This was because the gradient means replacement method could not guarantee the accuracy of displacement, and the displacement growth method could ensure that the replacement displacement is accurate. By comparing the experimental results with two kinds of displacement prediction methods for seed RF line, the paper found out that the neighborhood correlation coefficient method as a predictive method was better than that of the direct neighbor method. This was because the neighborhood correlation coefficient method could ensure the accuracy of the judgment during the process of determining the mean value of the displacement prediction, thus avoid the spread of the erroneous displacement and improve the accuracy of displacement estimation.