摘要
本文运用图像纹理的频谱特性提取特征参量，对Ｂ超图像进行纹理统计分析，能够诊断人体内部组织正常与否，并能确定病变之不同种类，为目前Ｂ超图像的定性观察上升到定量分析提供了一种新的诊断手段。
In this paper,We use several spectral features to statistically analyze the texture of the B-mode ultrasonic imagus of human liver in order to find the optimal spectral features in the automatic hepatosis-diagonosing system,unlike the original methods which were hased on spacial domain.In our experinent,We have found that two of all our features which are under consideration,namely the power spectrum entropy and variance,have the least error probability in the classification of the various type of liver tissues including the nomal and those with different kinds of hepatosis,and therefore are suitable for autodiagnosis purpose.
出处
《中国医学影像技术》
CSCD
1994年第1期59-62,共4页
Chinese Journal of Medical Imaging Technology
关键词
频谱特征
空域
频域
超声波诊断
Spectral feature
Spacial domain
Feature extraction
Entropy
Variance
Spectral domain
