基于环形阵列的乳腺超声检测仿真及成像方法研究

Research on simulation and imaging method for breast ultrasound imaging based on ring arrays

目的:研究一种适用于环形阵列的乳腺超声仿真和成像方法,实现对数值体模超声扫查过程的三维快速仿真,并重建得到对应的声速分布图。方法:先使用k-Wave对环形阵列及乳腺体模进行三维仿真并获得超声反射波和透射波的波形数据。再采用改进的小波赤池信息量准则(wavelet Akaike information criterion,wavelet-AIC)法提取声波第一到达时间,采用最短路径法得到正向模型,采用将L2型正则化作为惩罚项的混合版本的最小平方QR分解(least squares QR-factorization,LSQR)法求解反问题,采用迭代更新初始声速分布的方法实现对体模的声速分布成像。结果:优化后的三维仿真方法相比优化前仿真时间缩短了约33%,空间占用减少了约98%;改进后的成像方法计算得到的声速分布图更加清晰且与实际声速分布情况更为接近,其与体模声速分布图之间的均方根误差、相关系数和结构相似性分别为10.90 m/s、0.856、0.836。结论:该方法提高了环形阵列超声成像三维仿真速度,且得到的声速分布图可以正确显示体模的结构和声速差异,为下一步采用实际获取信号进行成像奠定了技术基础。

Objective To develop a breast ultrasound simulation and imaging method applicable to ring arrays to realize 3D rapid simulation of the numerical phantom ultrasound sweeping process and reconstruct the corresponding sound velocity distribution maps.Methods k-Wave was used to simulate the ring arrays and breast phantom in three dimensions and to obtain the waveform data of ultrasound reflected and transmitted waves.The improved wavelet Akaike information criterion(wavelet-AIC)method was used to extract the first arrival time of the acoustic wave,the forward model was obtained by the shortest path faster algorithm,the inverse problem was solved by the least squares QR-factorization(LSQR)method with L2-type regularization as a hybrid version of the penalty term,and the sound velocity distribution of the phantom was imaged by iteratively updating the initial sound velocity distribution.Results The optimized 3D simulation method shortened the simulation time by about 33%and reduced the space occupation by about 98%when compared with the pre-optimization method;the sound velocity distribution map by the improved imaging method was clear and close to the actual situations,which had the root mean square error,correlation coefficient and structural similarity with the phantom sound velocity distribution being 10.90 m/s,0.856 and 0.836,respectively.Conclusion The velocity for 3D simulation of ring array ultrasound imaging is improved,and the obtained sound velocity distribution map can correctly show the structure and sound velocity difference of the body phantom,which lays a technical foundation for imaging using actually acquired signals.