基于超声射频时间序列的肝纤维化程度评分方法

发布时间：2018-04-08 08:07

本文选题：超声射频时间序列　切入点：相对偏差评分法　出处：《生物医学工程学杂志》2015年01期

【摘要】：提出了基于超声射频(RF)时间序列的肝纤维化程度评分方法,尝试为肝纤维化的动态监测和诊断提供定量无创的影像辅助手段。首先使用宽频超声线阵探头获取活体肝组织的回波RF信号多帧;显示某帧B型图,选取感兴趣区(ROI)内的每点取其所有帧RF信号构成超声RF时间序列。提取ROI内RF时间序列的SMR(size measure relationship)分形维数均值和6个谱特征参数,分别以特征参数的相对偏差值和Fisher判别率作作相应特征的权值,对7个特征进行加权求和,得到评分Score-rd和Score-fisher。对两评分分别采用支持向量机(SVM)分类正确率和ROC曲线下面积(AUC)来评估其区分正常和肝硬化样本的有效性,结果显示Score-rd和Score-fisher的AUC依次为0.843、0.816,SVM分类正确率均为87.5%。说明提出的评分方法能够有效区分正常和肝硬化组织,有望为肝纤维化程度的诊断提供定量参考。
[Abstract]:A method for evaluating the degree of hepatic fibrosis based on the time series of radiofrequency ultrasound (RF) was proposed to provide quantitative and noninvasive imaging aids for the dynamic monitoring and diagnosis of hepatic fibrosis.Firstly, wideband ultrasonic linear array probe is used to obtain multiple frames of echo RF signals from living liver tissue, and each frame RF signal of each point in the region of interest is selected to form ultrasonic RF time series.The mean fractal dimension of SMR(size measure correlation of RF time series in ROI and six spectral characteristic parameters were extracted. The relative deviation of characteristic parameters and the Fisher discriminant rate were taken as the weights of the corresponding features respectively. The weighted sum of the seven features was carried out, and the scores of Score-rd and Score-fisherwere obtained.Support vector machine (SVM) classification accuracy rate and area under ROC curve were used to evaluate the validity of the two scores in differentiating normal and cirrhosis samples. The results showed that the AUC of Score-rd and Score-fisher were 0.843 / 0.816 respectively and 87.5% respectively.The results show that the proposed scoring method can effectively distinguish normal and liver cirrhosis tissues and may provide a quantitative reference for the diagnosis of liver fibrosis.
【作者单位】：华南理工大学电子与信息学院;中山大学附属肿瘤医院;
【基金】：国家自然科学基金资助项目(81271578)
【分类号】：R445.1;R575.2

【参考文献】