摘要
超声诊断除了受主观性和诊断经验等人为因素影响外,检测频率、机器参数的设置等成像条件的变化也会对图像产生很大的影响,因而使肝脏纤维化的量化分析受到很大限制。本文基于任意的肝脏超声图像提出了利用纹理边缘共生矩阵(TECM)进行纤维化量化分析的方法,首先由Canny算子提取纹理边缘,并计算其共生矩阵的熵作为分类的特征值。通过用leave one out最近邻法和Fisher线性分类器进行分类试验,证明其分类精度优于分形维(FD)。而且,用Fisher线性分类器对TECM和差分盒计数法(DBC)组成的联合特征向量(J DT)进行分类试验,当阈值为时,分类正确率(CCR)可以达到95. 1%;取阈值为时,灵敏性可以达到100%。
Ultrasound diagnosis has a lot of restricted application in liver fibrosis quantitative analysis, because ultrasound image conditions such as scanning frequency, machine settings can influence the ultrasound images besides its subjective and experience-based characteristic. This paper describes a novel feature extraction algorithm which is based on entropy of Texture edge co-occurrence matrix (TECM) from random ultrasound images. The edge co-occurrence matrix is computed by the texture edges, which is extracted by canny edge detector. TECM and FD are used to be developed and compared by both the Minimum Distance classification (by Leave-one-out scheme) and Fisher linear classification. Experiments on a set of two different tissue types show that the classification accuracy of TECM is better than that of FD. Based on the joint feature vector (J-DT), which is composed of both TECM and differential box-counting (DBC), the correct classification rate can reach 95.1% by Fisher linear classifier with threshold and the sensitivity can get 100% with threshold, respectively.
出处
《声学技术》
CSCD
2004年第F11期98-104,共7页
Technical Acoustics
关键词
超声
纤维化
纹理
边缘共生矩阵
Ultrasound
Fibrosis
Texture
Edge co-occurrence matrix
