摘要
目的根据乳腺超声图像的分级标准(BI-RADS)为诊断的指导,初步完成了诊断系统的设计。在图像处理中引入LBM滤波,并用无初始化的C-V模型进行分割,从形态特征与纹理特征入手,提取图像中相应的特征参数。采用支持向量机方法对所提取的特征参数进行分类。通过对88幅乳腺超声图像(其中良性37例、恶性51例)进行训练和测试,得到的判别准确率、敏感性和特异性分别为91.4%、94.4%和86.4%。结果表明,依据BI-RADS的分级特征研究有利于计算机辅助诊断在临床中的应用。
Based on an standard classification （ BI - RADS） used in breast ultrasound image, we design this computer - aided system. The lattice Boltzmann models is used here as an effective method to de - noise the speckle, also an improved CV models is adopted here, shape and texture features is both considered in extracting feature parameters. In the end support vector machines is used for classification. Experiments on 88 cases （ including 37 benign tumors and 51 malignant ones） achieve the accuracy91.4% , sensitivity94.4% and specificity 86.4%. Therefore, the result proves that this method based on BI - RADS can be helpful for computer - aided diagnosis.
出处
《生物医学工程学进展》
CAS
2009年第1期9-13,共5页
Progress in Biomedical Engineering
基金
上海市教育委员会科研创新项目资助(09YZ15)
关键词
乳腺超声图像
计算机辅助诊断
BI—RADS
LBM
C—V模型
SVM
Breast ultrasound image, Computer- aided diagnosis, BI- RADS, Lattice Boltzmann models, C - V models, Support vector machines
