摘要
文章首先研究了肝脏M型超声图像和B型超声图像的特征提取,提取了肝脏M型超声图像的运动包络、相隔一定时间的A型扫描线对的相关特征系数以及肝脏B型超声图像的共25个特征参数。在已提取的特征参数中,用特征融合的方法筛选出对肝硬化最敏感的特征组,并利用这20个参数的Fisher判据建立了基于超声图像特征的肝硬化无损自动分析系统。对临床43例的应用表明:该定量分析系统对肝硬化的诊断具有良好的效果。
Feature extraction of the liver's M-mode and B-mode ultrasonic images is first studied. A total of 25 features of liver motion envelope, the correlation coefficients between pairs of A-scan signals and the B-mode ultrasonic images are extracted. Useful features are then extracted using a feature fusion method. With 20 features chosen as such, an automatic non-invasive analysis system for hepatocirrhosis is constructed using the Fisher criterion. Based on the clinical applications of 43 cases, it is found that this quantitative analysis is effective in the diagnosis of hepatocirrhosis.
出处
《声学技术》
CSCD
2003年第4期288-291,共4页
Technical Acoustics
基金
上海市教委重点学科基金资助(KB20100)
