一种广义旁瓣相消的超声成像算法

针对最小方差和相干系数结合算法在超声成像对比度、分辨率和对噪声的鲁棒性方面存在的不足, 提出一种广义旁瓣相消的成像算法. 首先基于最小方差准则, 构造广义旁瓣相消器, 获得自适应与非自适应两部分加权向量, 然后根据接收信号的协方差矩阵构建特征阈值信号子空间, 并将获得的加权向量投影到此特征信号子空间的左奇异矢量空间中, 获得新的加权矢量. 通过Field II点目标和吸声斑仿真结果表明: 该方法获得的超声图像在对比度、分辨率和对噪声的鲁棒性上均优于传统延时叠加算法及最小方差和相干系数结合算法, 同时将广义旁瓣相消算法获得的加权向量与符号相干系数结合, 还能进一步提高超声图像质量. 最后通过geabr_0实验数据进行测试, 结果表明: 提出算法的分辨率和对比度及对噪声的鲁棒性上均优于传统延时叠加算法及最小方差和相干系数结合算法.

Ultrasound imaging algorithm based on generalized sidelobe canceller

The traditional delay and sum (DAS) algorithm is the most widely adopted method in medical ultrasound imaging; although it can produce images quickly, it sacrifices the resolution and the contrast ratio. The adaptive method such as the minimum variance (MV) continuously updates the apodization weighting vectors according to the received signals, so that the variance of the weighted signals is minimized, and thus the quality of the ultrasound imaging can be improved, especially its resolution. Although the image quality may be improved in the contrast ratio as well as the resolution after combining the minimum variance with the coherence factor (MV-CF), it complicates the algorithm, and the robustness against noise is enhanced but a little. An improved ultrasound imaging algorithm based on the generalized side lobe canceller (GSC) is proposed, which is constructed according to the minimum variance principle. The canceller is designed to classify the signal into desired and noise signals, combined with wiping off the big interferential eigenvectors, so that the robustness against noise can be enhanced. Firstly, the canceller divides the weighting vector into non-adaptive and adaptive weights, then the eigenstructure subspace is established according to the covariance matrix of the received signals, and the renewed weighting vector is achieved finally by projecting the weighting vector into the left singular space of the eigenstructure subspace. Simulations of the point targets and the cyst phantom through the simulation tool Field II demonstrate that the ultrasound image acquired through the proposed method is better than the traditional DAS and MV-CF algorithms in terms of the contrast ratio and resolution. In practice, the contrast ratio increases by roughly 7 dB compared to DAS and 5 dB to MV-CF. Furthermore, the proposed method gives a more satisfactory lateral resolution as well as the lowest side lobe peak level. From the sound-absorbing speckle simulation, the contrast ratio increases by 3 dB more than that of DAS and over 4 dB than that of MV-CF when noise exists. In addition, MV-CF performs the worst in the robustness aspect while the proposed GSC method makes improvement on the basis of it. Besides, the image quality can be further improved by combining the proposed method with sign coherence factor (GSC-SCF). After such a combination, the noise added to the data sets is almost invisible in point targets simulation. It also possesses the maximum mean power in cyst region in sound-absorbing speckle simulation. Finally, an experiment is conducted on the basis of the complete data sets which are offered by the University of Michigan. Results indicate that the proposed methods can perform better than the conventional DAS and MV-CF in resolution, contrast ratio and the robustness against noise.