基于相位信息的乳腺超声图像水平集分割  

Level set segmentation based on phase information of the breast ultrasound image

作　　者：陈雨羲[1] 刘奇[1] 黄韫栀[1] 张劲[1] 何凌[1] 邓丽华[1] 

机构地区：[1]四川大学医学信息工程系,四川成都610065

出　　处：《中国医学影像技术》2015年第3期463-466,共4页Chinese Journal of Medical Imaging Technology

基　　金：西华大学四川省信号与信息处理重点实验室基金项目(SGXZD0101-10-1)

摘　　要：目的基于相位信息改进距离正规化水平集演化(DRLSE)模型的速度收敛项,改善对乳腺肿瘤超声图像的分割效果。方法首先,利用Log-Gabor滤波器组对图像进行频域滤波,得到一组基于相位信息的特征图。其次,在相位一致性的基础上,求出乳腺超声图像经高斯噪声补偿后的最大方向能量相位PC(M),并采用细节保留各向异性扩散滤波(DPAD)模型对PC(M)降噪,减少斑点噪声的干扰。最后,选用Sigmoid函数,将滤波后的PC(M)作为其自变量,以替换DRLSE模型中的速度收敛项。结果采用改进后的模型对多幅临床乳腺肿瘤超声图像进行分割,分割结果显示基于相位信息的正规化水平集演化(PB-DRLSE)模型在相似性(SI)、真阳性(TP)和假阴性(FN)方面均优于原始DRLSE模型(P均<0.05)。结论本研究提出的分割方法较之原始模型对乳腺肿瘤超声图像的分割更为优越。Objective To improve the segmentation results of ultrasound images of breast tumor based on convergence speed embarking from the phase imformation of distance regularized level set evolution（DRLSE）model.Methods Firstly,images were filtrated using Log-Gabor filter bank in the frequency domain.Secondly,the direction of the maximum energy phase PC（M）of the breast ultrasound image after guassian noise compensation was obtained based on phase consistency.And the detail preserving anisotropic diffusion（DPAD）anisotropic diffusion model was used to reduce speckle noise interference for PC（M）.Finally,Sigmoid function was selected with the filtered PC（M）as independent variable to replace the convergence speed of DRLSE model.Results Compared with the original model,the segmentation results of phase basedDRLSE had preponderance（all P〈0.05）in similarity index（SI）,true positive（TP）and false negative（FN）.Conclusion The proposed segmentation method is superior to the original model in processing breast tumor ultrasound image.

关 键 词：超声检查 乳腺 LOG-GABOR 相位一致性 水平集 

分 类 号：R445.1[医药卫生—影像医学与核医学] R737.9[医药卫生—诊断学]